New Bengen Article Citing Kitces. SWR vs Valuation
New Bengen Article Citing Kitces. SWR vs Valuation
The bottom line is that the 4% rule holds. Nothing really new here except that Bengen reconciles his work with Kitces work.
"In summary, based on the earlier work of Michael Kitces, I have presented a tabular method to select an initial withdrawal rate for retirement portfolios, based on both recent inflation and stock market valuations. It exhibits a wide range of choices, between the “worst case” of 4.5% and a high of 13%, representing the full range of historically successful withdrawal rates. It is simple to use, though right now it applies only to taxadvantaged portfolios with a desired longevity of 30 years."
https://www.famag.com/news/choosingth ... section=40
"In summary, based on the earlier work of Michael Kitces, I have presented a tabular method to select an initial withdrawal rate for retirement portfolios, based on both recent inflation and stock market valuations. It exhibits a wide range of choices, between the “worst case” of 4.5% and a high of 13%, representing the full range of historically successful withdrawal rates. It is simple to use, though right now it applies only to taxadvantaged portfolios with a desired longevity of 30 years."
https://www.famag.com/news/choosingth ... section=40
Re: New Bengen Article Citing Kitces. SWR vs Valuation
Although with any Bengen reference, I have to remind readers that even he let emotions interfere. It seems that in 2008 he sold his clients out of equities/market timing. It is not clear when he got them back in, but one would assume that adventure into market timing was not positive:
http://www.intwealthmgt.com/files/3Bloo ... eFaith.pdf
http://www.intwealthmgt.com/files/3Bloo ... eFaith.pdf
Re: New Bengen Article Citing Kitces. SWR vs Valuation
Another good quote from the article:
I agree that SORR is often portrayed as the big risk for retirees to be concerned about. Inflation remains a big wild card for me. It's hard to know when it will "come back" (for some people it never left  think rising college or medical costs). Hard to know if the Fed/govt intervention can keep general inflation from getting too high. Hard to know if insuring against it (like with TIPS) is worth it. To deal with it, I'm depending on stocks to keep growing at a high enough rate. Failing that, working longer and spending less.I can’t emphasize enough the importance of inflation to a portfolio’s sustainability. The time line of inflation that a retiree experiences when they stop working is as important to their portfolio’s endurance as the sequence of investment returns, which has received much greater attention. This truth has been masked by a history of generally low inflation in the United States over the last 100 years (leaving out the 1970s and the period from 1948 to 1952).
Re: New Bengen Article Citing Kitces. SWR vs Valuation
Refreshing read. It is impressive to see how a rational mind approaches a rational problem. His particular attention to Valuation and Inflation are very appealing to me. As a recent retiree, attention to risk management in portfolio allocation and withdrawal formulation is critical, and he clearly approaches his analysis with these factors in line. Thanks for bring the paper to our (my) attention. I now have it in my online library.
Re: New Bengen Article Citing Kitces. SWR vs Valuation
If you consider returns in real, rather than nominal terms, doesn't that capture both affects (poor market performance and inflation)?sycamore wrote: ↑Wed Sep 30, 2020 10:51 am Another good quote from the article:I agree that SORR is often portrayed as the big risk for retirees to be concerned about. Inflation remains a big wild card for me. It's hard to know when it will "come back" (for some people it never left  think rising college or medical costs). Hard to know if the Fed/govt intervention can keep general inflation from getting too high. Hard to know if insuring against it (like with TIPS) is worth it. To deal with it, I'm depending on stocks to keep growing at a high enough rate. Failing that, working longer and spending less.I can’t emphasize enough the importance of inflation to a portfolio’s sustainability. The time line of inflation that a retiree experiences when they stop working is as important to their portfolio’s endurance as the sequence of investment returns, which has received much greater attention. This truth has been masked by a history of generally low inflation in the United States over the last 100 years (leaving out the 1970s and the period from 1948 to 1952).
Once in a while you get shown the light, in the strangest of places if you look at it right.
Re: New Bengen Article Citing Kitces. SWR vs Valuation
Thanks for sharing the article. I agree it is an interesting read. While I find it useful in understanding historical withdrawal rates under different scenarios of equity valuations (CAPE ratio), this is only part of the analysis. 50% of Bengen's article is in Intermediate Treasuries. With the 7year Treasury at 0.45% which have never occurred in the U.S. before. Bonds are starting at a much higher valuation than any prior historical period. That may or may not affect the Safe Max withdrawal rate if inflation remains low but it almost certainly would affect it if we see a spike in inflation.

 Posts: 3618
 Joined: Fri Aug 06, 2010 3:42 pm
Re: New Bengen Article Citing Kitces. SWR vs Valuation
The most important part of the Bengen article is the last 2 sentences IMO.
I hope we do get great returns going forward but I'm not counting on it. We're going to get a vaccine before long and that will help reverse a lot of the Covid19 induced damage to the economy, but much of that good future news is already priced into the market IMO. If we could only go backwards in time it would all be so easy.
Garland Whizzer
Backtesting is by its nature inherently flawed as an accurate predictor of the future. These safe withdrawal rates certainly work going backward, but will they work going forward from where we are now? I think the answer is: uncertain. I believe it best, given current circumstances, not to count on optimistic historically based safe withdrawal rates. Currently we have zero/negative real interest rates, generous equity valuations, persistent economic stagnation in all developed markets worldwide for more than a decade due to secular forces like debt and demographics that aren't going away any time soon. Add to that the current wide Covid19 gulf between exuberant stock prices and severely damaged economies with massive unemployment a lot of which will likely turn out the permanent not temporary. Equity markets in all markets worldwide are currently utterly dependent on extremely aggressive monetary policy plus massive fiscal policy debtfueled stimulus to keep them afloat. The entire corporate bond market in the US would have crashed without the FED's stepping up recently to buy BBB corporate bonds, something that was necessary but would have been unthinkable in times past. That is a measure of how vulnerable we are now. All this economic stimulus thrown in to the economy and yet it's a bit like beating a dead horse.I should also issue the usual cheerful disclaimer that this research is based on the analysis of historical data, and its application to future situations involves risk, as the future may differ significantly from the past. The term “safe” is meaningful only in its historical context, and does not imply a guarantee of future applicability.
I hope we do get great returns going forward but I'm not counting on it. We're going to get a vaccine before long and that will help reverse a lot of the Covid19 induced damage to the economy, but much of that good future news is already priced into the market IMO. If we could only go backwards in time it would all be so easy.
Garland Whizzer
 willthrill81
 Posts: 32250
 Joined: Thu Jan 26, 2017 2:17 pm
 Location: USA
 Contact:
Re: New Bengen Article Citing Kitces. SWR vs Valuation
The link you provided says nothing about Bengen timing anything. He suggested that clients reduce their withdrawals by 5 to 10 percent. That's not market timing.Leesbro63 wrote: ↑Wed Sep 30, 2020 9:53 am Although with any Bengen reference, I have to remind readers that even he let emotions interfere. It seems that in 2008 he sold his clients out of equities/market timing. It is not clear when he got them back in, but one would assume that adventure into market timing was not positive:
http://www.intwealthmgt.com/files/3Bloo ... eFaith.pdf
William Bengen, author of "Conserving Client Portfolios During Retirement,'' suggests that retirees reduce the amount they withdraw from their investment portfolios by 5 percent to 10 percent. If this turns out to be a long recession, "portfolio returns may be well below normal for several years,'' he says. Cutting back is an insurance policy against that possibility.
The Sensible Steward
Re: New Bengen Article Citing Kitces. SWR vs Valuation
Page 2, about the 3rd line up from the bottom:willthrill81 wrote: ↑Wed Sep 30, 2020 2:03 pmThe link you provided says nothing about Bengen timing anything. He suggested that clients reduce their withdrawals by 5 to 10 percent. That's not market timing.Leesbro63 wrote: ↑Wed Sep 30, 2020 9:53 am Although with any Bengen reference, I have to remind readers that even he let emotions interfere. It seems that in 2008 he sold his clients out of equities/market timing. It is not clear when he got them back in, but one would assume that adventure into market timing was not positive:
http://www.intwealthmgt.com/files/3Bloo ... eFaith.pdf
William Bengen, author of "Conserving Client Portfolios During Retirement,'' suggests that retirees reduce the amount they withdraw from their investment portfolios by 5 percent to 10 percent. If this turns out to be a long recession, "portfolio returns may be well below normal for several years,'' he says. Cutting back is an insurance policy against that possibility.
"Bengen also is keeping his clients out of equities. Normally, he says, he believes in traditional asset
allocation, but ``this is one of those rare instances when duckandcover is appropriate.'' "
I can find no other articles with follow up. Based on the market action since 20089, odds are that he got his clients back into "traditional asset allocation" at prices higher than when he sold.
Re: New Bengen Article Citing Kitces. SWR vs Valuation
So factoring all of that in, what do YOU think is a safe 30year SWR for a 50/50 portfolio?garlandwhizzer wrote: ↑Wed Sep 30, 2020 1:17 pm The most important part of the Bengen article is the last 2 sentences IMO.
Backtesting is by its nature inherently flawed as an accurate predictor of the future. These safe withdrawal rates certainly work going backward, but will they work going forward from where we are now? I think the answer is: uncertain. I believe it best, given current circumstances, not to count on optimistic historically based safe withdrawal rates. Currently we have zero/negative real interest rates, generous equity valuations, persistent economic stagnation in all developed markets worldwide for more than a decade due to secular forces like debt and demographics that aren't going away any time soon. Add to that the current wide Covid19 gulf between exuberant stock prices and severely damaged economies with massive unemployment a lot of which will likely turn out the permanent not temporary. Equity markets in all markets worldwide are currently utterly dependent on extremely aggressive monetary policy plus massive fiscal policy debtfueled stimulus to keep them afloat. The entire corporate bond market in the US would have crashed without the FED's stepping up recently to buy BBB corporate bonds, something that was necessary but would have been unthinkable in times past. That is a measure of how vulnerable we are now. All this economic stimulus thrown in to the economy and yet it's a bit like beating a dead horse.I should also issue the usual cheerful disclaimer that this research is based on the analysis of historical data, and its application to future situations involves risk, as the future may differ significantly from the past. The term “safe” is meaningful only in its historical context, and does not imply a guarantee of future applicability.
I hope we do get great returns going forward but I'm not counting on it. We're going to get a vaccine before long and that will help reverse a lot of the Covid19 induced damage to the economy, but much of that good future news is already priced into the market IMO. If we could only go backwards in time it would all be so easy.
Garland Whizzer
 willthrill81
 Posts: 32250
 Joined: Thu Jan 26, 2017 2:17 pm
 Location: USA
 Contact:
Re: New Bengen Article Citing Kitces. SWR vs Valuation
I didn't catch that.Leesbro63 wrote: ↑Wed Sep 30, 2020 2:08 pmPage 2, about the 3rd line up from the bottom:willthrill81 wrote: ↑Wed Sep 30, 2020 2:03 pmThe link you provided says nothing about Bengen timing anything. He suggested that clients reduce their withdrawals by 5 to 10 percent. That's not market timing.Leesbro63 wrote: ↑Wed Sep 30, 2020 9:53 am Although with any Bengen reference, I have to remind readers that even he let emotions interfere. It seems that in 2008 he sold his clients out of equities/market timing. It is not clear when he got them back in, but one would assume that adventure into market timing was not positive:
http://www.intwealthmgt.com/files/3Bloo ... eFaith.pdf
William Bengen, author of "Conserving Client Portfolios During Retirement,'' suggests that retirees reduce the amount they withdraw from their investment portfolios by 5 percent to 10 percent. If this turns out to be a long recession, "portfolio returns may be well below normal for several years,'' he says. Cutting back is an insurance policy against that possibility.
"Bengen also is keeping his clients out of equities. Normally, he says, he believes in traditional asset
allocation, but ``this is one of those rare instances when duckandcover is appropriate.'' "
I can find no other articles with follow up. Based on the market action since 20089, odds are that he got his clients back into "traditional asset allocation" at prices higher than when he sold.
Many here 'blinked' and sold in 2008. I don't doubt that a lot of Bengen's clients were insistent on him doing that. Regardless, that doesn't change my views on any of his research, which many others have consequently replicated.
The Sensible Steward
 willthrill81
 Posts: 32250
 Joined: Thu Jan 26, 2017 2:17 pm
 Location: USA
 Contact:
Re: New Bengen Article Citing Kitces. SWR vs Valuation
Considering that real bond returns were about 1.6% from 19411981, I don't see future bond returns looking any worse than that.
Stocks are valued more highly than they have been historically, and CAPE has very strongly correlated with 30 year SWRs. That said, CAPE alone does not suggest that the '4% rule of thumb' is in danger.
Nonetheless, no one should be using a SWR method in the first place, and I don't believe that anyone actually does. Everyone makes adjustments to their withdrawals, especially when their portfolio is performing poorly, and this is perfectly rational.
The Sensible Steward
Re: New Bengen Article Citing Kitces. SWR vs Valuation
It's unfortunate that Bengen is dancing around the real elephant in the room, which isn't "inflation regime" directly (as Bengen claims) but rather the expected REAL return of bonds at the beginning of the period. Expected inflation is irrelevant to the SWR calculation, yet Bengen focuses on a metric (trailing 12month CPI) that is essentially a measure of expected inflation.gjlynch17 wrote: ↑Wed Sep 30, 2020 1:07 pm Thanks for sharing the article. I agree it is an interesting read. While I find it useful in understanding historical withdrawal rates under different scenarios of equity valuations (CAPE ratio), this is only part of the analysis. 50% of Bengen's article is in Intermediate Treasuries. With the 7year Treasury at 0.45% which have never occurred in the U.S. before. Bonds are starting at a much higher valuation than any prior historical period. That may or may not affect the Safe Max withdrawal rate if inflation remains low but it almost certainly would affect it if we see a spike in inflation.
Being fair to Bengen, he can't easily run his SAFEMAX backtests using direct measures of expected real returns. It's still disheartening to see such fundamental errors from a widely respected author.
"Far more money has been lost by investors preparing for corrections than has been lost in corrections themselves." ~~ Peter Lynch
 willthrill81
 Posts: 32250
 Joined: Thu Jan 26, 2017 2:17 pm
 Location: USA
 Contact:
Re: New Bengen Article Citing Kitces. SWR vs Valuation
Bonds had real returns of about 1.6% from 19411981, so negative real bond yields are far from unprecedented. And the '4% rule of thumb' held up throughout that period (well, 3.8% in 1966).
The Sensible Steward
 Uncorrelated
 Posts: 1237
 Joined: Sun Oct 13, 2019 3:16 pm
Re: New Bengen Article Citing Kitces. SWR vs Valuation
I would argue that using SWR also signifies big problems. At this point is should be obvious that it is possible to obtain much higher SWR's with a dynamic asset allocation. It should also be obvious that real investors don't pick a withdrawal rate and stick to that no matter what. If that wasn't enough, the work is full of data snooping biases.
I can't take these researchers seriously anymore. Poor models, unrealistic assumptions, statistically insignificant conclusions, overfitting and weak optimization methods are everywhere.
I can't take these researchers seriously anymore. Poor models, unrealistic assumptions, statistically insignificant conclusions, overfitting and weak optimization methods are everywhere.
Re: New Bengen Article Citing Kitces. SWR vs Valuation
So if you had $1M at age 65, how much would you spend in the first year?Uncorrelated wrote: ↑Wed Sep 30, 2020 2:50 pm I would argue that using SWR also signifies big problems. At this point is should be obvious that it is possible to obtain much higher SWR's with a dynamic asset allocation. It should also be obvious that real investors don't pick a withdrawal rate and stick to that no matter what. If that wasn't enough, the work is full of data snooping biases.
I can't take these researchers seriously anymore. Poor models, unrealistic assumptions, statistically insignificant conclusions, overfitting and weak optimization methods are everywhere.
Re: New Bengen Article Citing Kitces. SWR vs Valuation
I’m not Garland, but I think the 4% number has help up well over the last 3 decades.
I would roll in very low bond returns (negative) and a low inflation regime (positive) to leave that number unchanged.
It's not an engineering problem  Hersh Shefrin  To get the "risk premium", you really do have to take the risk  nisiprius
 Uncorrelated
 Posts: 1237
 Joined: Sun Oct 13, 2019 3:16 pm
Re: New Bengen Article Citing Kitces. SWR vs Valuation
I'm not sure, but I would either implement this paper to find out, or use aiplanner to make an estimate. Although it looks like a black box, I understand the models moderately well, the recommended withdrawal in the first year is 50k plus social security income.Leesbro63 wrote: ↑Wed Sep 30, 2020 2:59 pmSo if you had $1M at age 65, how much would you spend in the first year?Uncorrelated wrote: ↑Wed Sep 30, 2020 2:50 pm I would argue that using SWR also signifies big problems. At this point is should be obvious that it is possible to obtain much higher SWR's with a dynamic asset allocation. It should also be obvious that real investors don't pick a withdrawal rate and stick to that no matter what. If that wasn't enough, the work is full of data snooping biases.
I can't take these researchers seriously anymore. Poor models, unrealistic assumptions, statistically insignificant conclusions, overfitting and weak optimization methods are everywhere.
Re: New Bengen Article Citing Kitces. SWR vs Valuation
So you seem to be confirming, as did Bengen and Kitces here, that the (at least) 4% "rule of thumb" is fine. ??Uncorrelated wrote: ↑Wed Sep 30, 2020 4:03 pmI'm not sure, but I would either implement this paper to find out, or use aiplanner to make an estimate. Although it looks like a black box, I understand the models moderately well, the recommended withdrawal in the first year is 50k plus social security income.Leesbro63 wrote: ↑Wed Sep 30, 2020 2:59 pmSo if you had $1M at age 65, how much would you spend in the first year?Uncorrelated wrote: ↑Wed Sep 30, 2020 2:50 pm I would argue that using SWR also signifies big problems. At this point is should be obvious that it is possible to obtain much higher SWR's with a dynamic asset allocation. It should also be obvious that real investors don't pick a withdrawal rate and stick to that no matter what. If that wasn't enough, the work is full of data snooping biases.
I can't take these researchers seriously anymore. Poor models, unrealistic assumptions, statistically insignificant conclusions, overfitting and weak optimization methods are everywhere.
Re: New Bengen Article Citing Kitces. SWR vs Valuation
Not Uncorrelated, but for real world example, age 65 was not the first year, but at 65 I had planned to withdraw about 5.3% of the initial portfolio value but actually withdrew about 1%. The year before the plan estimated about 4.6% and the actual was about 3%. In the first years of retirement before starting Social Security the withdrawals have ranged from 0.5% to 3.7% and after starting Social Security from 0.0% to 2.7%. Annual spending in retirement has ranged from half the average spending to twice the average spending.Leesbro63 wrote: ↑Wed Sep 30, 2020 2:59 pmSo if you had $1M at age 65, how much would you spend in the first year?Uncorrelated wrote: ↑Wed Sep 30, 2020 2:50 pm I would argue that using SWR also signifies big problems. At this point is should be obvious that it is possible to obtain much higher SWR's with a dynamic asset allocation. It should also be obvious that real investors don't pick a withdrawal rate and stick to that no matter what. If that wasn't enough, the work is full of data snooping biases.
I can't take these researchers seriously anymore. Poor models, unrealistic assumptions, statistically insignificant conclusions, overfitting and weak optimization methods are everywhere.

 Posts: 16882
 Joined: Fri Apr 10, 2015 12:29 am
Re: New Bengen Article Citing Kitces. SWR vs Valuation
What would be the advantage of trying to estimate and use expected real returns when simulating retirement outcomes by backtesting? We have the actual returns and actual inflation for the periods of time used, and that is what Bengen used.vineviz wrote: Being fair to Bengen, he can't easily run his SAFEMAX backtests using direct measures of expected real returns. It's still disheartening to see such fundamental errors from a widely respected author.
Re: New Bengen Article Citing Kitces. SWR vs Valuation
Northern Flicker wrote: ↑Wed Sep 30, 2020 4:52 pmWhat would be the advantage of trying to estimate and use expected real returns when simulating retirement outcomes by backtesting? We have the actual returns and actual inflation for the periods of time used, and that is what Bengen used.vineviz wrote: Being fair to Bengen, he can't easily run his SAFEMAX backtests using direct measures of expected real returns. It's still disheartening to see such fundamental errors from a widely respected author.
Not speaking for vineviz, just my thought.
If there were a reasonable correlation between expected returns and actual returns (I am skeptical about equities, but maybe more useful for bonds?), then that might allow linking the stratified back test results to the path one might expect going forward.
Once in a while you get shown the light, in the strangest of places if you look at it right.
Re: New Bengen Article Citing Kitces. SWR vs Valuation
The goal should be to control for, as much as possible, the expected success of the withdrawal strategy at the beginning of the historical periods we are examining. That's the only way to evaluate whether the expected success of the same strategy today is more or less likely that it was in the past.Northern Flicker wrote: ↑Wed Sep 30, 2020 4:52 pmWhat would be the advantage of trying to estimate and use expected real returns when simulating retirement outcomes by backtesting? We have the actual returns and actual inflation for the periods of time used, and that is what Bengen used.vineviz wrote: Being fair to Bengen, he can't easily run his SAFEMAX backtests using direct measures of expected real returns. It's still disheartening to see such fundamental errors from a widely respected author.
Historically, the most challenged cohort of retirees was the group retiring in the mid1960s. I typically use the 1966 cohort as the worst case, so let's take them as an example.
The starting conditions in January, 1966 are these: inflation over the prior decade has been a little under 2%, so let's use that as our expected inflation rate. 5year Treasuries are yielding 4.88% at the end of December 1965. Add a 5% equity risk premium to that yield as the expected return for stocks. Given those three parameters, using a Monte Carlo simulation what is the probability that a 4% withdrawal rate (adjusted for inflation), will succeed? Answer: about 93% chance of success. In other words, even though this group got an unlucky (10th percentile) draw in terms of level of returns and sequence of returns the 4% withdrawal rate worked.
Try the same thing using today's starting numbers: 2% inflation, 0.24% yield on 5year Treasury, and the same 5% equity risk premium. What's the current probability of "success" with a 4% SWR? Answer: 43%.
So why is this exercise important? Because the 1966 scenario (which is historical the worst SWR starting date) just barely worked with a 10th percentile outcome. If today's retirees are equally as unlucky as the 1966 group (i.e. they get a 10th percentile outcome using TODAY'S starting conditions), they'll be out of money in 19 years using the 4% SWR instead of 30 years.
Clearly there is room to debate the assumptions I used (should we smooth the Treasury yield instead of taking the instantaneous value, should we adjust the expected ERP to account for the higher CAPE today, etc.), but any attempt to extrapolate past success to the present situation NEEDS to control for the different independent variables.
"Far more money has been lost by investors preparing for corrections than has been lost in corrections themselves." ~~ Peter Lynch
Re: New Bengen Article Citing Kitces. SWR vs Valuation
could you please explain dynamic AA, perhaps with an example?Uncorrelated wrote: ↑Wed Sep 30, 2020 2:50 pm I would argue that using SWR also signifies big problems. At this point is should be obvious that it is possible to obtain much higher SWR's with a dynamic asset allocation. It should also be obvious that real investors don't pick a withdrawal rate and stick to that no matter what. If that wasn't enough, the work is full of data snooping biases.
I can't take these researchers seriously anymore. Poor models, unrealistic assumptions, statistically insignificant conclusions, overfitting and weak optimization methods are everywhere.
Thanks
 willthrill81
 Posts: 32250
 Joined: Thu Jan 26, 2017 2:17 pm
 Location: USA
 Contact:
Re: New Bengen Article Citing Kitces. SWR vs Valuation
I thought that you did not care much for data prior to the Volcker era. If that is the case, then I don't see why we would worry about interest rates going up significantly to rein in high inflation (i.e. real bond returns may be slightly negative over the next decade but aren't likely to be anywhere near as bad as those from 19771981).vineviz wrote: ↑Wed Sep 30, 2020 7:26 pmThe goal should be to control for, as much as possible, the expected success of the withdrawal strategy at the beginning of the historical periods we are examining. That's the only way to evaluate whether the expected success of the same strategy today is more or less likely that it was in the past.Northern Flicker wrote: ↑Wed Sep 30, 2020 4:52 pmWhat would be the advantage of trying to estimate and use expected real returns when simulating retirement outcomes by backtesting? We have the actual returns and actual inflation for the periods of time used, and that is what Bengen used.vineviz wrote: Being fair to Bengen, he can't easily run his SAFEMAX backtests using direct measures of expected real returns. It's still disheartening to see such fundamental errors from a widely respected author.
Historically, the most challenged cohort of retirees was the group retiring in the mid1960s. I typically use the 1966 cohort as the worst case, so let's take them as an example.
The starting conditions in January, 1966 are these: inflation over the prior decade has been a little under 2%, so let's use that as our expected inflation rate. 5year Treasuries are yielding 4.88% at the end of December 1965. Add a 5% equity risk premium to that yield as the expected return for stocks. Given those three parameters, using a Monte Carlo simulation what is the probability that a 4% withdrawal rate (adjusted for inflation), will succeed? Answer: about 93% chance of success. In other words, even though this group got an unlucky (10th percentile) draw in terms of level of returns and sequence of returns the 4% withdrawal rate worked.
Try the same thing using today's starting numbers: 2% inflation, 0.24% yield on 5year Treasury, and the same 5% equity risk premium. What's the current probability of "success" with a 4% SWR? Answer: 43%.
So why is this exercise important? Because the 1966 scenario (which is historical the worst SWR starting date) just barely worked with a 10th percentile outcome. If today's retirees are equally as unlucky as the 1966 group (i.e. they get a 10th percentile outcome using TODAY'S starting conditions), they'll be out of money in 19 years using the 4% SWR instead of 30 years.
Clearly there is room to debate the assumptions I used (should we smooth the Treasury yield instead of taking the instantaneous value, should we adjust the expected ERP to account for the higher CAPE today, etc.), but any attempt to extrapolate past success to the present situation NEEDS to control for the different independent variables.
Another issue I see is whether it's plausible to expect intermediateterm Treasuries to only return .24% over the entirety of the next 30 years. Historically, starting yields of intermediateterm bonds have not been predictive of the next 30 years' returns.
Finally, I'm still quite skeptical of the MC simulation approach for estimating forward success. Using the approach you've laid out, what were the odds that the '4% rule' would hold up for every 30 year period, as it actually did (with the 3.8% exception for 1966 retirees)? My guess is well below 1%. Maybe U.S. retirees really did get that lucky. But I suspect that it's more likely that the method does not capture reality.
The Sensible Steward

 Posts: 16882
 Joined: Fri Apr 10, 2015 12:29 am
Re: New Bengen Article Citing Kitces. SWR vs Valuation
As long as you want to use (estimated) expected real returns, you might as well use CPIE (CPI for elderly Americans) or some measure of elder American's personal inflation rate.vineviz wrote: The goal should be to control for, as much as possible, the expected success of the withdrawal strategy at the beginning of the historical periods we are examining. That's the only way to evaluate whether the expected success of the same strategy today is more or less likely that it was in the past.
Re: New Bengen Article Citing Kitces. SWR vs Valuation
These analyses are missing two very crucial factors: low bond yields (discussed upthread) and possible scientific breakthroughs leading to substantially increased life expectancy for the investing class that reaches retirement.
I've never seen any financial planning "experts" discuss the latter probability. When I've raised the issue before, it has been suggested to me that if such a thing were to happen, surely there must have been robust equity growth due to advancing technology. But increases in technological advancement don't have to equate to good equity returns (citation: sinceinception investment returns of airline and nuclearpowerrelated stocks).
It would seem that the solution to the aforementioned problem would be to work longer, but when you're 100, you may no longer be able to physically work in a meaningful way and almost certainly will not be to go back in time and change decisions made at 60. This is doubly true for those who are trying to retire in their 30s on portfolios needing a 34% WR to meet their expenses.
I think a groundup variable withdrawal framework planning for a longer life expectancy would be more reasonable than the SWR framework.
I've never seen any financial planning "experts" discuss the latter probability. When I've raised the issue before, it has been suggested to me that if such a thing were to happen, surely there must have been robust equity growth due to advancing technology. But increases in technological advancement don't have to equate to good equity returns (citation: sinceinception investment returns of airline and nuclearpowerrelated stocks).
It would seem that the solution to the aforementioned problem would be to work longer, but when you're 100, you may no longer be able to physically work in a meaningful way and almost certainly will not be to go back in time and change decisions made at 60. This is doubly true for those who are trying to retire in their 30s on portfolios needing a 34% WR to meet their expenses.
I think a groundup variable withdrawal framework planning for a longer life expectancy would be more reasonable than the SWR framework.
 Uncorrelated
 Posts: 1237
 Joined: Sun Oct 13, 2019 3:16 pm
Re: New Bengen Article Citing Kitces. SWR vs Valuation
Sure. If you follow the asset allocation detailed in this figure, you will have the highest possible success rate with the given modeling assumptions:1210sda wrote: ↑Wed Sep 30, 2020 8:07 pmcould you please explain dynamic AA, perhaps with an example?Uncorrelated wrote: ↑Wed Sep 30, 2020 2:50 pm I would argue that using SWR also signifies big problems. At this point is should be obvious that it is possible to obtain much higher SWR's with a dynamic asset allocation. It should also be obvious that real investors don't pick a withdrawal rate and stick to that no matter what. If that wasn't enough, the work is full of data snooping biases.
I can't take these researchers seriously anymore. Poor models, unrealistic assumptions, statistically insignificant conclusions, overfitting and weak optimization methods are everywhere.
Thanks
The color indicates the asset allocation (percentage allocated to bonds). The X axis indicates the number of years left until the end of your 30year retirement period. The Y axis indicates your current ratio between net worth and spending.
If you follow a 4% spending rule and 30 year retirement, then in the first year you should have a balanced allocation (approx 50/50, dead center in the image). If the market does poorly, the ratio of capital to spending goes down and you'll end up in the red zone, where the stock allocation is increased. If the market does well, you'll end up in the blue zone where a more conservative AA is applicable. You should rebalance your portfolio each year according to this figure. The contour lines indicate the confidence thresholds of a successful retirement. With this asset allocation, the historical SWR is approx 4.5% compared to approx 3.8% for traditional approaches. source here.
(standard disclaimer: do not use this asset allocation: high risk of behavioral errors, does not support stochastic lifespan, does not support variable spending. Although this is a big improvement over previous SWR literature, there are fundamental errors that can only be fixed by abandoning the SWR metric altogether. Read Gordon Irlam's research for asset allocations that fix all these problems.)

 Posts: 2055
 Joined: Fri Jun 07, 2019 2:00 am
 Location: Florida
Re: New Bengen Article Citing Kitces. SWR vs Valuation
+1dbr wrote: ↑Wed Sep 30, 2020 4:39 pmNot Uncorrelated, but for real world example, age 65 was not the first year, but at 65 I had planned to withdraw about 5.3% of the initial portfolio value but actually withdrew about 1%. The year before the plan estimated about 4.6% and the actual was about 3%. In the first years of retirement before starting Social Security the withdrawals have ranged from 0.5% to 3.7% and after starting Social Security from 0.0% to 2.7%. Annual spending in retirement has ranged from half the average spending to twice the average spending.Leesbro63 wrote: ↑Wed Sep 30, 2020 2:59 pmSo if you had $1M at age 65, how much would you spend in the first year?Uncorrelated wrote: ↑Wed Sep 30, 2020 2:50 pm I would argue that using SWR also signifies big problems. At this point is should be obvious that it is possible to obtain much higher SWR's with a dynamic asset allocation. It should also be obvious that real investors don't pick a withdrawal rate and stick to that no matter what. If that wasn't enough, the work is full of data snooping biases.
I can't take these researchers seriously anymore. Poor models, unrealistic assumptions, statistically insignificant conclusions, overfitting and weak optimization methods are everywhere.
One spends and withdraws according to necessities that fluctuate from year to year. The total income that is available from what one has accumulated plus any additional income vs. liabilities will dictate the spending on an annual basis. Who really needs a formula (SWR) to figure out this basic math? Practicality dictates that one cannot spend more than they can afford to. It's simple to understand and easy to implement...no complexity necessary.
"Success is going from failure to failure without loss of enthusiasm." Winston Churchill.
Re: New Bengen Article Citing Kitces. SWR vs Valuation
Yes, there are definitely limits to the comparability of data from prior periods to today, and I think we should be sensitive to that. I guess what I'm trying to nudge people towards is to focus on base rates.willthrill81 wrote: ↑Wed Sep 30, 2020 8:33 pm I thought that you did not care much for data prior to the Volcker era. If that is the case, then I don't see why we would worry about interest rates going up significantly to rein in high inflation (i.e. real bond returns may be slightly negative over the next decade but aren't likely to be anywhere near as bad as those from 19771981).
I'm not a fan of sports games, so pardon me if this analogy is off, but there's a basketball player named Stephen Curry who I think is pretty good. From what I can tell, he makes about 56% of the shots he takes overall. But it'd be unwise to assume that is his probability of landing the shot IF WE KNOW what kind of shot he's taking. Why? Because he makes 91% of his free throw attempts and "only" 43% of his threepoint attempts. Retirees in 1966 were like Steph Curry shooting a free throw. Retirees in 2020 are like Steph Curry shooting from behind the threepoint line.
That's fair. I agree the analysis I did is somewhat sensitive to the assumptions I use, though not THAT sensitive: swap out the 0.24% ITT with a 1.2% LTT and the odds of success only go up to 48% from 43%: the real yield curve doesn't look great for investors at any maturity. I used 5year yields in my example because it's roughly the "riskfree" rate to which the equity premiums is applied, five years is about the upper limit for accurately forecasting inflation, and because it's about the average duration of outstanding debt.willthrill81 wrote: ↑Wed Sep 30, 2020 8:33 pmAnother issue I see is whether it's plausible to expect intermediateterm Treasuries to only return .24% over the entirety of the next 30 years. Historically, starting yields of intermediateterm bonds have not been predictive of the next 30 years' returns.
"Far more money has been lost by investors preparing for corrections than has been lost in corrections themselves." ~~ Peter Lynch
Re: New Bengen Article Citing Kitces. SWR vs Valuation
The purpose of SWR estimation is to help the household plan their retirement spending. At the most simplified level, a dollar you spend in Year One of retirement is a dollar you can NOT spend in Year Two or Year Twenty.rossington wrote: ↑Thu Oct 01, 2020 4:26 am Who really needs a formula (SWR) to figure out this basic math? Practicality dictates that one cannot spend more than they can afford to. It's simple to understand and easy to implement...no complexity necessary.
The purpose of the SWR is to help households benchmark how much they CAN spend without compromising variation in their standard of living. As you say, it's not a rule for determine how much you MUST spend: that's impacted by the need and desire to consume as well.
As long as the NEED to spend is less than the ABILITY to spend, retirement income planning is relatively easy. When the DESIRE to spend is greater than the ABILITY to spend, some choices must be made. When the NEED to spend is greater than the ABILITY to spend, a retiree finds themselves between a rock and hard place which is why so much energy is consumed by planning to avoid this scenario.
"Far more money has been lost by investors preparing for corrections than has been lost in corrections themselves." ~~ Peter Lynch
Re: New Bengen Article Citing Kitces. SWR vs Valuation
I don't trust any study not considering current low expected returns for both bonds and stocks.
If you plug in expected returns from Vanguard for a 60/40 portfolio to a monte carlo simulation, you get around 2.50% Perpetual Withdrawal Rate in the median sample. This totally ignores sequence of returns risk, it's just the number that would retain portfolio value if all assets return the expected number every year. I'd start from here for long retirement periods.
I don't trust monte carlo simulations for tail cases (since no mean reversion), but i think the median simulation means something.
If you plug in expected returns from Vanguard for a 60/40 portfolio to a monte carlo simulation, you get around 2.50% Perpetual Withdrawal Rate in the median sample. This totally ignores sequence of returns risk, it's just the number that would retain portfolio value if all assets return the expected number every year. I'd start from here for long retirement periods.
I don't trust monte carlo simulations for tail cases (since no mean reversion), but i think the median simulation means something.
Re: New Bengen Article Citing Kitces. SWR vs Valuation
This isn't a general characteristic of Monte Carlo tools. In fact, it's pretty common for Monte Carlo simulators to incorporate mean reversion fucntionality. And inside the 90% or 95% confidence interval, it's probably not a big deal in most cases.
"Far more money has been lost by investors preparing for corrections than has been lost in corrections themselves." ~~ Peter Lynch
Re: New Bengen Article Citing Kitces. SWR vs Valuation
ok. same 60/40 portfolio then gives 3.24% SWR for 30y at 10p.
For, 40y it becomes 2.50% (same number i got for 10y PWR at 50p).
With GARCH model, it improves to 2.63%. I guess GARCH model includes mean reversion.
 willthrill81
 Posts: 32250
 Joined: Thu Jan 26, 2017 2:17 pm
 Location: USA
 Contact:
Re: New Bengen Article Citing Kitces. SWR vs Valuation
I agree that the starting yields of bonds and the expected returns of stocks certainly matter, but I'm not convinced that they matter as much as your analysis implies. There were many periods of the past where the starting real yields of bonds were much lower than they are now, yet the '4% rule' held up throughout all of them. The same goes for stocks, though we admittedly don't have many instances where valuations were as high as they are now.vineviz wrote: ↑Thu Oct 01, 2020 6:33 amYes, there are definitely limits to the comparability of data from prior periods to today, and I think we should be sensitive to that. I guess what I'm trying to nudge people towards is to focus on base rates.willthrill81 wrote: ↑Wed Sep 30, 2020 8:33 pm I thought that you did not care much for data prior to the Volcker era. If that is the case, then I don't see why we would worry about interest rates going up significantly to rein in high inflation (i.e. real bond returns may be slightly negative over the next decade but aren't likely to be anywhere near as bad as those from 19771981).
I'm not a fan of sports games, so pardon me if this analogy is off, but there's a basketball player named Stephen Curry who I think is pretty good. From what I can tell, he makes about 56% of the shots he takes overall. But it'd be unwise to assume that is his probability of landing the shot IF WE KNOW what kind of shot he's taking. Why? Because he makes 91% of his free throw attempts and "only" 43% of his threepoint attempts. Retirees in 1966 were like Steph Curry shooting a free throw. Retirees in 2020 are like Steph Curry shooting from behind the threepoint line.
That's fair. I agree the analysis I did is somewhat sensitive to the assumptions I use, though not THAT sensitive: swap out the 0.24% ITT with a 1.2% LTT and the odds of success only go up to 48% from 43%: the real yield curve doesn't look great for investors at any maturity. I used 5year yields in my example because it's roughly the "riskfree" rate to which the equity premiums is applied, five years is about the upper limit for accurately forecasting inflation, and because it's about the average duration of outstanding debt.willthrill81 wrote: ↑Wed Sep 30, 2020 8:33 pmAnother issue I see is whether it's plausible to expect intermediateterm Treasuries to only return .24% over the entirety of the next 30 years. Historically, starting yields of intermediateterm bonds have not been predictive of the next 30 years' returns.
I wish you had addressed my big issue with your analysis, which I outlined below.
willthrill81 wrote: ↑Wed Sep 30, 2020 8:33 pm Finally, I'm still quite skeptical of the MC simulation approach for estimating forward success. Using the approach you've laid out, what were the odds that the '4% rule' would hold up for every 30 year period, as it actually did (with the 3.8% exception for 1966 retirees)? My guess is well below 1%. Maybe U.S. retirees really did get that lucky. But I suspect that it's more likely that the method does not capture reality.
The Sensible Steward

 Posts: 4428
 Joined: Mon May 26, 2008 10:20 am
 Location: Second star on the right and straight on 'til morning
Re: New Bengen Article Citing Kitces. SWR vs Valuation
We live in hope. Also, might be one answer to your earlier question about SWR on a 50/50 portfolio.Leesbro63 wrote: ↑Wed Sep 30, 2020 9:32 am The bottom line is that the 4% rule holds.
https://www.famag.com/news/choosingth ... section=40
FI is the best revenge. LBYM. Invest the rest. Stay the course. Die anyway.  PS: The cavalry isn't coming, kids. You are on your own.
Re: New Bengen Article Citing Kitces. SWR vs Valuation
1) This isn't true, certainly not in the modern era (e.g. 1950s to present). From 1956 to 2008, there was never a negative expected real yield on 5year Treasuries and the average expected real yield over that period was 2.3%.willthrill81 wrote: ↑Thu Oct 01, 2020 9:16 am There were many periods of the past where the starting real yields of bonds were much lower than they are now, yet the '4% rule' held up throughout all of them.
In short, our current real yield is in the bottom 5% of years since 1955 and the current CAPE ratio on the S&P 500 is in the top 10%. The combination is, literally, the least favorable in modern history.
Statistically speaking, I just don't find that criticism very powerful. We already know that the 4% rule worked in all the past periods: the probability that it did so is 100%. Calculating the conditional probability of a 4% SWR working for each of the possible 30year periods in history would be laborious, and I don't think it would give us any additional information we could use to make decisions now.willthrill81 wrote: ↑Thu Oct 01, 2020 9:16 am I wish you had addressed my big issue with your analysis, which I outlined below.
"Far more money has been lost by investors preparing for corrections than has been lost in corrections themselves." ~~ Peter Lynch
Re: New Bengen Article Citing Kitces. SWR vs Valuation
I think the point he is making is that your statistical model may not be accurately capturing what actually happens under particular starting conditions.vineviz wrote: ↑Thu Oct 01, 2020 10:56 am Statistically speaking, I just don't find that criticism very powerful. We already know that the 4% rule worked in all the past periods: the probability that it did so is 100%. Calculating the conditional probability of a 4% SWR working for each of the possible 30year periods in history would be laborious, and I don't think it would give us any additional information we could use to make decisions now.
The model says current conditions have high probability of failure. What does the model say about previous starting conditions. if it also predicts that many of them had a very high probability of failure, yet they all succeeded, then either we got really lucky, repeatedly, or one has to consider the possibility that the model may be faulty.
Once in a while you get shown the light, in the strangest of places if you look at it right.

 Posts: 16882
 Joined: Fri Apr 10, 2015 12:29 am
Re: New Bengen Article Citing Kitces. SWR vs Valuation
I think it is very difficult to estimate expected inflation and expected return at points of time in history. One way the research could be formulated to take differing conditions into account would be to analyze the success rste of withdrawals as an increment over the riskfree rate.
Another point is that if managing assets for longevity, ie to minimize the probability of running out of assets, then it is likely that annuitizing one's allocation to nominal bonds always reduces that probability.
If one's asset allocation were say 50% stock, 25% treasuries/nominal bonds, 25% TIPS, then annuitizing the 25% in nominal bonds in retirement probably makes sense. A caveat is that this reduces rebalancing opportunities.
Another point is that if managing assets for longevity, ie to minimize the probability of running out of assets, then it is likely that annuitizing one's allocation to nominal bonds always reduces that probability.
If one's asset allocation were say 50% stock, 25% treasuries/nominal bonds, 25% TIPS, then annuitizing the 25% in nominal bonds in retirement probably makes sense. A caveat is that this reduces rebalancing opportunities.
 willthrill81
 Posts: 32250
 Joined: Thu Jan 26, 2017 2:17 pm
 Location: USA
 Contact:
Re: New Bengen Article Citing Kitces. SWR vs Valuation
There were several years in the 1940s when there were, TMK. I don't see the purpose in acting as though those didn't exist.vineviz wrote: ↑Thu Oct 01, 2020 10:56 am1) This isn't true, certainly not in the modern era (e.g. 1950s to present).willthrill81 wrote: ↑Thu Oct 01, 2020 9:16 am There were many periods of the past where the starting real yields of bonds were much lower than they are now, yet the '4% rule' held up throughout all of them.
Precisely. vineviz has put forth what the model's predicted likelihood of failure for 1966 retirees was, which is fine even though we know that they succeeded. Now, I want to see a comprehensive backtest of the model, including the a priori likelihood of zero failures occurring over all periods examined, which is a valid test. If the model would have predicted a 5% chance of failure in any given year (for the sake of illustration), the likelihood of no failures occurring over 50 periods would only be 7.7%. Maybe the U.S. really did get that lucky. But I view it as being much more likely that the model is flawed.marcopolo wrote: ↑Thu Oct 01, 2020 11:54 amI think the point he is making is that your statistical model may not be accurately capturing what actually happens under particular starting conditions.vineviz wrote: ↑Thu Oct 01, 2020 10:56 am Statistically speaking, I just don't find that criticism very powerful. We already know that the 4% rule worked in all the past periods: the probability that it did so is 100%. Calculating the conditional probability of a 4% SWR working for each of the possible 30year periods in history would be laborious, and I don't think it would give us any additional information we could use to make decisions now.
The model says current conditions have high probability of failure. What does the model say about previous starting conditions. if it also predicts that many of them had a very high probability of failure, yet they all succeeded, then either we got really lucky, repeatedly, or one has to consider the possibility that the model may be faulty.
The Sensible Steward

 Posts: 3618
 Joined: Fri Aug 06, 2010 3:42 pm
Re: New Bengen Article Citing Kitces. SWR vs Valuation
In terms of nominal return I believe bonds will provide between 1% and 2% nominal over the next 30 years. For real returns, the kind that matter for those who must purchase things in real time, we subtract inflation which again I suspect will be between 1% and 2%. In short I expect bonds to provide stability to a portfolio but not to produce any significant real returns going forward for a long, long time. We're at the end of a fabulous 4 decade long record setting bond bull market in which 10 yr Treasury rates reduced from 15% to 0.6%. Now we'll start paying for that wonderful ride. Going forward I expect the opposite, a long struggle to make any positive real gains. Rates can certainly drop from here but it's hard to get rich on negative nominal yields. So my guess and that's all it is, a guess, is that half of a 50/50 portfolio will produce no significant difference from zero real return.leesbro63 wrote:
So factoring all of that in, what do YOU think is a safe 30year SWR for a 50/50 portfolio?
As for equites, there is a much wider range of outcomes. My rather conservative formula is divided rate plus the inverse of PE. The current dividend rate of TSM is 1.5%. Assuming that remains stable going forward, a rather big and likely faulty assumption, that adds 1.5% real to stock returns going forward. We add to the dividend the estimated price appreciation in the TSM index itself which I rather conservatively estimate at the inverse of the current PE ratio (27.5) which would add another 3.6% in expected price appreciation. That adds up to 5.1% assuming PE ratios and dividend yields remain stable, huge assumptions. So if we assume zero from bonds and 5.1% from stocks a 50/50 portfolio using these admittedly flawed assumptions is expected to produce a 2.5%  2.6% real long term portfolio return. These estimates have such a wide range of errors in either direction, up or down. Like expected return projections done by numerous experts published a decade or so ago, they may appear totally laughable in retrospect. They are conservatively based which is I believe appropriate.
There is so much money concentrated now in the investing class that financial assets like stocks, bonds, real estate, gold, etc., may continue to be bid up higher and higher in price inflating risk returns much more. PE ratios in the 30s may become the norm or we may revert to lower PE but we're very unlike to return to historical averages of about 16. We could face black swans, progressively increasing inflation which no one expects, or major geopolitical, climate, or natural disaster upheavals that would have serious adverse effects on stocks. Bonds come in handy when stocks tank so even if they don't produce real returns, bonds can be life savers.
These predictions, like all predictions no matter who makes them are to be taken lightly. Using current parameters they derive the conservative central base case projection but there is a wide range of potential outcomes on either side. Good luck.
Garland Whizzer
 willthrill81
 Posts: 32250
 Joined: Thu Jan 26, 2017 2:17 pm
 Location: USA
 Contact:
Re: New Bengen Article Citing Kitces. SWR vs Valuation
In this thread, I showed a mathematically valid means of doing so with the time value of money formula.
Using CAPE, the expected real return of U.S. stocks is 3.2% over the next decade. The yield on 10 year TIPS is .98%. So the expected real return for a 60/40 portfolio over the next decade is 1.53%. Assuming a 30 year withdrawal period and no bequest (i.e. an ending portfolio value of zero), the first year's withdrawal would be $41,186.
The Sensible Steward
Re: New Bengen Article Citing Kitces. SWR vs Valuation
That’s the thing, though: when you account for previous starting conditions then what you see is that the times a 4% withdrawal rate worked in the past had a high ex ante probability of working.marcopolo wrote: ↑Thu Oct 01, 2020 11:54 am What does the model say about previous starting conditions. if it also predicts that many of them had a very high probability of failure, yet they all succeeded, then either we got really lucky, repeatedly, or one has to consider the possibility that the model may be faulty.
The fact that it worked well in the past isn’t remarkable, in other words.
"Far more money has been lost by investors preparing for corrections than has been lost in corrections themselves." ~~ Peter Lynch
Re: New Bengen Article Citing Kitces. SWR vs Valuation
So we’re back to 4%.willthrill81 wrote: ↑Thu Oct 01, 2020 2:25 pmIn this thread, I showed a mathematically valid means of doing so with the time value of money formula.
Using CAPE, the expected real return of U.S. stocks is 3.2% over the next decade. The yield on 10 year TIPS is .98%. So the expected real return for a 60/40 portfolio over the next decade is 1.53%. Assuming a 30 year withdrawal period and no bequest (i.e. an ending portfolio value of zero), the first year's withdrawal would be $41,186.
Re: New Bengen Article Citing Kitces. SWR vs Valuation
big difference is this formula is dynamic. it's not constant dollars. it's not even constant percentage. you may get $20k in the second year.Leesbro63 wrote: ↑Thu Oct 01, 2020 2:38 pmSo we’re back to 4%.willthrill81 wrote: ↑Thu Oct 01, 2020 2:25 pmIn this thread, I showed a mathematically valid means of doing so with the time value of money formula.
Using CAPE, the expected real return of U.S. stocks is 3.2% over the next decade. The yield on 10 year TIPS is .98%. So the expected real return for a 60/40 portfolio over the next decade is 1.53%. Assuming a 30 year withdrawal period and no bequest (i.e. an ending portfolio value of zero), the first year's withdrawal would be $41,186.
The formula i intend to use is SWR=AVERAGE(3% , CAPEYield) = 3.11% (as of today)
3% is the expected real return of my portfolio based on Vanguard projections.
Re: New Bengen Article Citing Kitces. SWR vs Valuation
So, it would seem that it again comes down to sequence of returms. Your conservative estimate is about double what is needed to sustain a 4% SWR, if the real returns are in a straight line. But, that can change, for better or worse, depending on the sequence that produce the same long term real return.garlandwhizzer wrote: ↑Thu Oct 01, 2020 2:18 pmIn terms of nominal return I believe bonds will provide between 1% and 2% nominal over the next 30 years. For real returns, the kind that matter for those who must purchase things in real time, we subtract inflation which again I suspect will be between 1% and 2%. In short I expect bonds to provide stability to a portfolio but not to produce any significant real returns going forward for a long, long time. We're at the end of a fabulous 4 decade long record setting bond bull market in which 10 yr Treasury rates reduced from 15% to 0.6%. Now we'll start paying for that wonderful ride. Going forward I expect the opposite, a long struggle to make any positive real gains. Rates can certainly drop from here but it's hard to get rich on negative nominal yields. So my guess and that's all it is, a guess, is that half of a 50/50 portfolio will produce no significant difference from zero real return.leesbro63 wrote:
So factoring all of that in, what do YOU think is a safe 30year SWR for a 50/50 portfolio?
As for equites, there is a much wider range of outcomes. My rather conservative formula is divided rate plus the inverse of PE. The current dividend rate of TSM is 1.5%. Assuming that remains stable going forward, a rather big and likely faulty assumption, that adds 1.5% real to stock returns going forward. We add to the dividend the estimated price appreciation in the TSM index itself which I rather conservatively estimate at the inverse of the current PE ratio (27.5) which would add another 3.6% in expected price appreciation. That adds up to 5.1% assuming PE ratios and dividend yields remain stable, huge assumptions. So if we assume zero from bonds and 5.1% from stocks a 50/50 portfolio using these admittedly flawed assumptions is expected to produce a 2.5%  2.6% real long term portfolio return. These estimates have such a wide range of errors in either direction, up or down. Like expected return projections done by numerous experts published a decade or so ago, they may appear totally laughable in retrospect. They are conservatively based which is I believe appropriate.
There is so much money concentrated now in the investing class that financial assets like stocks, bonds, real estate, gold, etc., may continue to be bid up higher and higher in price inflating risk returns much more. PE ratios in the 30s may become the norm or we may revert to lower PE but we're very unlike to return to historical averages of about 16. We could face black swans, progressively increasing inflation which no one expects, or major geopolitical, climate, or natural disaster upheavals that would have serious adverse effects on stocks. Bonds come in handy when stocks tank so even if they don't produce real returns, bonds can be life savers.
These predictions, like all predictions no matter who makes them are to be taken lightly. Using current parameters they derive the conservative central base case projection but there is a wide range of potential outcomes on either side. Good luck.
Garland Whizzer
Once in a while you get shown the light, in the strangest of places if you look at it right.
Re: New Bengen Article Citing Kitces. SWR vs Valuation
No matter how you cut it, we keep coming back to about 4%. As a practical matter, most people can never save 25x spending (to fund a 4% SWR), let alone higher amounts. It would seem that for most, a 4% initial withdrawal, adjusted somewhat for good and bad times, is a good rule of thumb.marcopolo wrote: ↑Thu Oct 01, 2020 3:12 pmSo, it would seem that it again comes down to sequence of returms. Your conservative estimate is about double what is needed to sustain a 4% SWR, if the real returns are in a straight line. But, that can change, for better or worse, depending on the sequence that produce the same long term real return.garlandwhizzer wrote: ↑Thu Oct 01, 2020 2:18 pmIn terms of nominal return I believe bonds will provide between 1% and 2% nominal over the next 30 years. For real returns, the kind that matter for those who must purchase things in real time, we subtract inflation which again I suspect will be between 1% and 2%. In short I expect bonds to provide stability to a portfolio but not to produce any significant real returns going forward for a long, long time. We're at the end of a fabulous 4 decade long record setting bond bull market in which 10 yr Treasury rates reduced from 15% to 0.6%. Now we'll start paying for that wonderful ride. Going forward I expect the opposite, a long struggle to make any positive real gains. Rates can certainly drop from here but it's hard to get rich on negative nominal yields. So my guess and that's all it is, a guess, is that half of a 50/50 portfolio will produce no significant difference from zero real return.leesbro63 wrote:
So factoring all of that in, what do YOU think is a safe 30year SWR for a 50/50 portfolio?
As for equites, there is a much wider range of outcomes. My rather conservative formula is divided rate plus the inverse of PE. The current dividend rate of TSM is 1.5%. Assuming that remains stable going forward, a rather big and likely faulty assumption, that adds 1.5% real to stock returns going forward. We add to the dividend the estimated price appreciation in the TSM index itself which I rather conservatively estimate at the inverse of the current PE ratio (27.5) which would add another 3.6% in expected price appreciation. That adds up to 5.1% assuming PE ratios and dividend yields remain stable, huge assumptions. So if we assume zero from bonds and 5.1% from stocks a 50/50 portfolio using these admittedly flawed assumptions is expected to produce a 2.5%  2.6% real long term portfolio return. These estimates have such a wide range of errors in either direction, up or down. Like expected return projections done by numerous experts published a decade or so ago, they may appear totally laughable in retrospect. They are conservatively based which is I believe appropriate.
There is so much money concentrated now in the investing class that financial assets like stocks, bonds, real estate, gold, etc., may continue to be bid up higher and higher in price inflating risk returns much more. PE ratios in the 30s may become the norm or we may revert to lower PE but we're very unlike to return to historical averages of about 16. We could face black swans, progressively increasing inflation which no one expects, or major geopolitical, climate, or natural disaster upheavals that would have serious adverse effects on stocks. Bonds come in handy when stocks tank so even if they don't produce real returns, bonds can be life savers.
These predictions, like all predictions no matter who makes them are to be taken lightly. Using current parameters they derive the conservative central base case projection but there is a wide range of potential outcomes on either side. Good luck.
Garland Whizzer
 willthrill81
 Posts: 32250
 Joined: Thu Jan 26, 2017 2:17 pm
 Location: USA
 Contact:
Re: New Bengen Article Citing Kitces. SWR vs Valuation
Yes, a different withdrawal will be calculated for each year. But it's extremely unlikely that the withdrawal would be cut in half from one year to the next, even if stocks dropped by 50%. A big part of the reason for this is because expected returns would increase if stocks' price fell. If you refer to the thread I linked to above, you'll see an example that covered the 20082009 bear market, and withdrawals across the period were quite consistent.klaus14 wrote: ↑Thu Oct 01, 2020 2:55 pmbig difference is this formula is dynamic. it's not constant dollars. it's not even constant percentage. you may get $20k in the second year.Leesbro63 wrote: ↑Thu Oct 01, 2020 2:38 pmSo we’re back to 4%.willthrill81 wrote: ↑Thu Oct 01, 2020 2:25 pmIn this thread, I showed a mathematically valid means of doing so with the time value of money formula.
Using CAPE, the expected real return of U.S. stocks is 3.2% over the next decade. The yield on 10 year TIPS is .98%. So the expected real return for a 60/40 portfolio over the next decade is 1.53%. Assuming a 30 year withdrawal period and no bequest (i.e. an ending portfolio value of zero), the first year's withdrawal would be $41,186.
The Sensible Steward
 willthrill81
 Posts: 32250
 Joined: Thu Jan 26, 2017 2:17 pm
 Location: USA
 Contact:
Re: New Bengen Article Citing Kitces. SWR vs Valuation
As far as starting withdrawals go, 4% has held up well historically, at least in the U.S., and it's true that more sophisticated approaches than the SWR method often arrive at a similar percentage for those in their mid 60s. Heck, even RMDs start at close to 4% (though they aren't withdrawals, just a tax event). I have not found any data, either current or historic, that lead to seriously question the viability of retirees around age 65 at least starting their withdrawals at around 4%.Leesbro63 wrote: ↑Thu Oct 01, 2020 3:26 pm No matter how you cut it, we keep coming back to about 4%. As a practical matter, most people can never save 25x spending (to fund a 4% SWR), let alone higher amounts. It would seem that for most, a 4% initial withdrawal, adjusted somewhat for good and bad times, is a good rule of thumb.
The Sensible Steward
Re: New Bengen Article Citing Kitces. SWR vs Valuation
willthrill81 wrote: ↑Thu Oct 01, 2020 4:52 pmYes, a different withdrawal will be calculated for each year. But it's extremely unlikely that the withdrawal would be cut in half from one year to the next, even if stocks dropped by 50%. A big part of the reason for this is because expected returns would increase if stocks' price fell. If you refer to the thread I linked to above, you'll see an example that covered the 20082009 bear market, and withdrawals across the period were quite consistent.klaus14 wrote: ↑Thu Oct 01, 2020 2:55 pmbig difference is this formula is dynamic. it's not constant dollars. it's not even constant percentage. you may get $20k in the second year.Leesbro63 wrote: ↑Thu Oct 01, 2020 2:38 pmSo we’re back to 4%.willthrill81 wrote: ↑Thu Oct 01, 2020 2:25 pmIn this thread, I showed a mathematically valid means of doing so with the time value of money formula.
Using CAPE, the expected real return of U.S. stocks is 3.2% over the next decade. The yield on 10 year TIPS is .98%. So the expected real return for a 60/40 portfolio over the next decade is 1.53%. Assuming a 30 year withdrawal period and no bequest (i.e. an ending portfolio value of zero), the first year's withdrawal would be $41,186.
You are forgetting to take into account that the "valuations matter" arguments presented in this forum seems to only ratchet in one direction. That is, the WR is always argued down when valuations rise, but none of those making that argument seem to come back and say the WR can rise when valuations drop. Where were all the "4% is the new 3%" posts when the market dropped 30% in March?
Once in a while you get shown the light, in the strangest of places if you look at it right.