I am reading another book (of course I am) that was recommended by Mr. B called "Beyond the 4% Rule" by Abraham Okusanya 2018. So it was written after Mr. B's 2016 update but before his 2025 book. Mr. Okusanya is the Mr. B of the UK as most of his data concerns the UK. Anyway, he states "The Data available on UK equity valuations is somewhat limited, going back to 1926 for PE ratio and 1935 for Shiller PE10 (CAPE ratio). The data shows clearly that there is an inverse relationship between the PE ratio in a given year and the annualized real return on equities over the subsequent 10 years. However, we should be extremely cautious if we use current equity valuations to predict future equity returns, in the medium to long term." He bases this on the R squared between current PE ratio and subsequent 10 years' real return on UK equities is 20% and 22.5% on CAPE. Meaning the predictability of future return (in the UK anyway) using current PE or CAPE is lower than the flip of a coin.
@pizzaman Have you experimented with Pralana's CAPE-based withdrawal system? It is based on a "best fit" line between cape and withdrawal, where the user sets the slope and intercept. (I use 1.75 and 0.5.) For reasons unknown to me, Pralana only allows us to use it with historical runs. But it seems potentially even more valuable looking forward ten years, for withdrawals and monte carlos.
@jkandell I am still using Gold 2025 so I have not used Pralana's CAPE-based withdrawal system. Be careful just using CAPE to determine safe withdraw rates (SWR). Inflation tided with CAPE is more predictive. CAPE hit 32.6% in 1929 which predicted the bear market of 1929-1932. However, even though the stock market dropped by almost 90%, the SAFEMAX for that era was 5.8% because of the deflationary backdrop of the period.
Ern’s safemax for 1929 peak was 4.79%. But if you augment with his “drop from peak”, he would have had you withdraw more each month as the stock market crash grew month by month till June 1932 (79% real drawdown) when he would have said to withdraw 5.26%. So ending up with similar result? But in the real world, I doubt if I’d been there I’d have the trust nor stomach to follow either author and raise my withdrawal rate during 1929-32!
Interesting. "...till June 1932 (79% real drawdown) when he would have said to withdraw 5.26%". Did Ern use inflation coupled with CAPE to get to the 5.26%? If he did not use inflation, then you could say that today, with CAPE greater than 30, your safe withdraw rate would also be about 5.26%?
ERN only uses inflation-adjusted figures for everything.
A) With the advantage of hindsight, the SWRs (in real dollars) for each 30 year cohorts 1929-32 turned out to have been as follows (with my approximation of Bengen's 11/11/11/11/40/5 portfolio in ERN's spreadsheet). It ranged from almost 5% to about 11% in retrospect. But who knew that at time in the depths of despair that swr would have been so high? (The worst years were of course the late 1960s cohorts, but who knew that either at the time?)
| date | year | SWR |
| 2/28/1929 | 1929 | 4.86% |
| 3/31/1929 | 1929 | 4.89% |
| 4/30/1929 | 1929 | 4.96% |
| 5/31/1929 | 1929 | 4.89% |
| 6/30/1929 | 1929 | 5.14% |
| 7/31/1929 | 1929 | 4.92% |
| 8/31/1929 | 1930 | 4.89% |
| 9/30/1929 | 1930 | 4.79% |
| 10/31/1929 | 1930 | 4.91% |
| 11/30/1929 | 1930 | 5.56% |
| 12/31/1929 | 1930 | 5.95% |
| 1/31/1930 | 1930 | 5.91% |
| 2/28/1930 | 1930 | 5.65% |
| 3/31/1930 | 1930 | 5.54% |
| 4/30/1930 | 1930 | 5.21% |
| 5/31/1930 | 1930 | 5.29% |
| 6/30/1930 | 1930 | 5.34% |
| 7/31/1930 | 1930 | 5.90% |
| 8/31/1930 | 1931 | 5.74% |
| 9/30/1930 | 1931 | 5.73% |
| 10/31/1930 | 1931 | 6.30% |
| 11/30/1930 | 1931 | 6.60% |
| 12/31/1930 | 1931 | 6.62% |
| 1/31/1931 | 1931 | 6.94% |
| 2/28/1931 | 1931 | 6.57% |
| 3/31/1931 | 1931 | 6.05% |
| 4/30/1931 | 1931 | 6.23% |
| 5/31/1931 | 1931 | 6.67% |
| 6/30/1931 | 1931 | 7.04% |
| 7/31/1931 | 1931 | 6.48% |
| 8/31/1931 | 1932 | 6.76% |
| 9/30/1931 | 1932 | 6.80% |
| 10/31/1931 | 1932 | 8.25% |
| 11/30/1931 | 1932 | 7.93% |
| 12/31/1931 | 1932 | 8.22% |
| 1/31/1932 | 1932 | 9.11% |
| 2/29/1932 | 1932 | 9.00% |
| 3/31/1932 | 1932 | 8.67% |
| 4/30/1932 | 1932 | 9.15% |
| 5/31/1932 | 1932 | 10.14% |
| 6/30/1932 | 1932 | 11.45% |
| 7/31/1932 | 1932 | 11.40% |
| 8/31/1932 | 1933 | 9.27% |
| 9/30/1932 | 1933 | 7.18% |
| 10/31/1932 | 1933 | 7.38% |
| 11/30/1932 | 1933 | 8.13% |
| 12/31/1932 | 1933 | 8.43% |
B) Now how does this same 1929-33 period look if using safemax with cape-regime filtering? (As if we’d teleported with erns spreadsheet back to the past.)
ERN uses a formula 0.5%*1/cape + 1.75% to adjust withdrawal to valuation. (Pralana has a version of this equation you can explore on the historical analysis page!)
If folks back then had been using a Cape-adjusted withdrawal rate, ERN in theory would have recommended the following SWRs for each monthly 30 cohort.
It looks like, because of high-cape, it would have started around 3.5% and gown down a little as CAPE peaked, but then come back up as CAPE dropped.
All this is hypothetical of course, since the 11/11/11/11 was not really able to be invested back then, and SAFEmax was taking into account the history that came after.
| Month | Year | Return | CAPE | Withdrawal Rate |
| 1 | 1929 | 2.08% | 27.79 | 3.55% |
| 2 | 1929 | -0.36% | 27.53 | 3.57% |
| 3 | 1929 | -1.07% | 26.83 | 3.61% |
| 4 | 1929 | 1.83% | 28.04 | 3.53% |
| 5 | 1929 | -4.55% | 26.57 | 3.63% |
| 6 | 1929 | 4.87% | 29.41 | 3.45% |
| 7 | 1929 | 0.82% | 30.43 | 3.39% |
| 8 | 1929 | 2.51% | 33.37 | 3.25% |
| 9 | 1929 | -2.13% | 31.46 | 3.34% |
| 10 | 1929 | -11.50% | 25.04 | 3.75% |
| 11 | 1929 | -6.17% | 21.59 | 4.07% |
| 12 | 1929 | 0.98% | 22.10 | 4.01% |
| 1 | 1930 | 5.17% | 23.49 | 3.88% |
| 2 | 1930 | 2.25% | 23.98 | 3.84% |
| 3 | 1930 | 6.82% | 25.87 | 3.68% |
| 4 | 1930 | -1.29% | 25.32 | 3.73% |
| 5 | 1930 | -0.58% | 24.81 | 3.77% |
| 6 | 1930 | -9.13% | 20.81 | 4.15% |
| 7 | 1930 | 3.30% | 21.70 | 4.05% |
| 8 | 1930 | 0.55% | 21.91 | 4.03% |
| 9 | 1930 | -8.77% | 18.84 | 4.40% |
| 10 | 1930 | -4.19% | 17.20 | 4.66% |
| 11 | 1930 | 0.18% | 17.10 | 4.67% |
| 12 | 1930 | -4.15% | 15.85 | 4.90% |
| 1 | 1931 | 6.23% | 16.79 | 4.73% |
| 2 | 1931 | 8.92% | 18.89 | 4.40% |
| 3 | 1931 | -2.35% | 17.64 | 4.58% |
| 4 | 1931 | -6.18% | 16.01 | 4.87% |
| 5 | 1931 | -4.81% | 13.29 | 5.51% |
| 6 | 1931 | 9.15% | 16.06 | 4.86% |
| 7 | 1931 | -3.75% | 15.11 | 5.06% |
| 8 | 1931 | -0.07% | 15.22 | 5.04% |
| 9 | 1931 | -17.23% | 10.70 | 6.42% |
| 10 | 1931 | 4.73% | 11.64 | 6.05% |
| 11 | 1931 | -3.01% | 10.60 | 6.47% |
| 12 | 1931 | -9.23% | 9.10 | 7.24% |
| 1 | 1932 | 1.93% | 9.37 | 7.08% |
| 2 | 1932 | 4.41% | 9.56 | 6.98% |
| 3 | 1932 | -4.56% | 8.47 | 7.65% |
| 4 | 1932 | -9.14% | 6.78 | 9.12% |
| 5 | 1932 | -10.80% | 5.27 | 11.24% |
| 6 | 1932 | 1.35% | 5.25 | 11.27% |
| 7 | 1932 | 24.01% | 7.22 | 8.67% |
| 8 | 1932 | 29.98% | 9.97 | 6.76% |
| 9 | 1932 | -2.32% | 9.68 | 6.91% |
| 10 | 1932 | -8.74% | 8.39 | 7.71% |
| 11 | 1932 | -2.97% | 7.96 | 8.03% |
| 12 | 1932 | 1.04% | 8.44 | 7.67% |
I don't think ERN would say you actually would use these exact numbers; I think he goes on to smooth them to 12 month averages. He was just testing his CAPE: SWR equation (similar to what Bengen does with his best fit line).
Thanks for starting the discussion. I admit that at first I am wondering why we need a discussion on "Safe Withdrawal Rates". The topic has been beat to death by researchers and the media.
I was interested in his work when I retired about 8 years ago because I did not use Pralana at the time. I read his paper on the 4% rule and decided to recreate his research to see if I could maybe look into other issues as well. It was a fun and geeky project and I was able to match his work. I used the same data and found that there is a very tight correlation to the first 10 years and the subsequent withdrawal rate. Others have also claimed that the first 5 to 10 years are critical but I never found any data to back it up. I used the data set to investigate other things like asset allocation and longevity as well.
The other insight from the data set was the fact that many of those retirees that started with a % withdrawal and increased it by inflation and never made any changes would need a strong stomach to watch their portfolio go to a very low level before starting to recover. I don't think many would be able to keep the course.
Then I started using Pralana to do my planning in a more methodical way. Now a safe withdrawal rate has no connection to my plan because my expenses change through time with the heavy spending on discretionary items for the first 15 years and then declines in the typical three phases of retirement.
I had become sick and tired of hearing about all the safe withdrawal research and wondered why anyone would want to use it and if anybody really did.
I have since, listened to a few interviews with Bill and a good Q&A. I realized that he is not focused on a magic number that we should all use. Instead, he has extended the research into many levels as you guys point out. For example, he thinks the withdrawal rate is dynamic and that we should take current market conditions into account to determine the starting withdrawal rate and then make changes as retirement and conditions change. He even advocates changing asset allocation through time as markets get more expensive or less expensive.
I also follow a "Safety First" retirement style and have all my non-discretionary expenses and most of my discretionary expenses covered by secure income. I don't use bond ladders because they can only last as long as the ladder period. That makes much of this type of work irrelevant to me. He even said in one of the interviews that secure income was not his focus of research but agrees that it could play a very important role for many retirees.
I just purchased his book and look forward to seeing what other insights he has developed using a broader data set. It will be more of an educational read just because I like this stuff. So thanks for the discussion.
@golich428 i too follow the “safety first” and have all my essentials covered by matching tips or ibonds etc. And i pmt what’s left, possibly similar to you. But i still find swr useful
But my aim is constant discretionary amount. So i use the SWR as a check for my discretionary money after taking into account my LMP. (BtW jeske ERN’s spreadsheet generates a “safe consumption rate”, not safe “withdrawal”, because it gives you a figure net of the cash flow of essentials and pensions/ss.)
But here’s the thing: If you run this kind of customized swr analysis every year, with changing portfolio and discount rates, the SCR will end up virtually the same as an actuarial style withdrawal. The SWR idea gets a lot of flack; but if re-run every year it becomes a defacto variable actuarial withdrawal method.
@pizzaman, from interviews i gather that bengen now based swr on safemax filtered for inflation regimes. From interviews he looks at (1) current cpi, (2) estimated 5 year inflation. He breaks inflation down into 2-5% (ordinary), >5% (high), <2% (low), with different safemaxes for each taking cape into account.
i believe your table 2.3 reflects the normal 2-5% inflation regimes?
It’s a really neat idea. There does seem something very different about 1970s vs 1929. Does he have a table of "high inflation" regimes and cape?
I have read about 80% of the book and I give him credit for digging into the data but he seems to over fit the data by using polynomials with a degree of 5. It seems to me that you could use his approach as a first pass to screen if you have enough to to retire but beyond that I would not rely on it. There is other free software like FI Calc that is easier to use. But in the end, using Pralana is still my preferred approach.
I do want to give him credit for explaining all the pitfalls with his approach and would recommend reading it if this stuff is of interest.
If u
@golich428 I like Mr. B's book a lot. I would not suggest that anybody rely solely on it, but it gives you a safe base to start with, and then you can modify to your heart's content 😉. Not sure what you mean by "...he seems to over fit the data by using polynomials with a degree of 5." Can I assume that you do not really follow a safe withdraw rate? Do you have an overreaching retirement planning philosophy? Just curious.
@jkandell Yes to Table 2.3 reflects the normal 2-5% inflation regimes.
Not sure what you mean by "...he seems to over fit the data by using polynomials with a degree of 5."
In data analysis you don't want to force the "best fit" curve that looks good on the surface (e.g. very high R2 with data) but isn't valid.
Polynomial fits are famous for this, because if your equation is complex enough (Bengen's) you'll get a real tight fit. But it isn't "real", for predictive purposes, if that makes sense. There is very little actual data here, and if he were more sophisticated with statistics he'd realize he is over-reaching with that level of precision.
That was the basis for my preference for the way Jeske fit the CAPE data in his analysis, with a simple linear regression line (not a polynomial), and lumping CAPE into broad buckets for broad generalizations, rather than finding the perfect equation.
No tables for other inflation regimes?
@pizzaman Jonathan summed it up - in my opinion he is fitting the noise. When I looked at this about 8 years ago I had a R^2 similar to Bill’s for the relationship between CAPE10 and I think SWR or S&P 500 (can’t remember which) using a log function but I knew it was just fiction as a predictor. It looked good though.
No I do not follow a safe withdrawal rate approach. I follow a Safety First + approach meaning I like secure income to cover almost all of my spending. I do this for two reasons: 1. I don’t need to rely on market returns to live the lifestyle I desire and 2. I suffer from the accumulation mindset that says my portfolio should always be increasing - this makes it much harder to pull money out of my portfolio for very discretionary stuff. I like my money being “pushed” into my spending account much more than I like having to “pull” money into my spending account. I know it is irrational but the behavior is real - just ask my wife.
I guess there is a third reason - my goal is to to do things I desire now and not put them off into the future because my safe withdrawal rule i use says I can only spend a fixed amount of money every year or have to increase it or decrease it based on some rule. Then wake up 10 years from now and look back and say: Well honey I guess we should have done those things because look we have way more now than we predicted but now we are not physically able to. The opposite could be true too.
I guess there is a fourth reason as well. Secure income is mentally easy - trying to monitor a safe withdrawal rate like Bill recommends when your mental capacity is declining would not be easy at all. Not many are immune to this problem but most of us think we are.
Many retirees cover there essentials with secure income - I just took the concept a step further - that is the reason it is +.
Now my portfolio can be sliced into reserves for long term care, unexpected expenses and legacy. Once my reserves are set aside, what is leftover can be used as I wish for extra discretionary items that seem to come up that I did not include in my plan. We don’t follow an annual budget - we are trying to do as much as possible while we are young and able.
I don’t explicitly use historical returns either. No doubt they inform the process of using forward looking capital market assumptions but as Bill points out current market conditions do matter. The YTM on bonds is a good predictor of future returns so if you are using a SWR based on a YTM = 10% or 1% when YTM is 4% then that could be a problem. If CAPE10 is approaching 40 and your SWR is based on a CAPE of 20 then that could be a problem. The CAPE10 relationship is noisy at best. Bill has included inflation as part of his update which I think is a step in the right direction. Others I think have looked at this too. These relationships are noisy so we can only expect to hit the target in retirement planning not the bullseye.
I also do not use current stock market conditions to change my equity allocation. I keep it the same and adjust my forward ROR. However, I do take profits after big run ups to add to my liquidity.
For bonds I do the same but I may adjust my duration based on the yield curve. For example, I am not going to hold a bond fund with a 6 year duration, if the YTM is not high enough to compensate for the added interest rate risk. We saw in 2022 that if you did you got burned. Bill calls this risk management and takes recommendations from a third party. Not my cup of tea but if it works for him then that is a good thing.
By the way, I liked Bill’s book too. He put in a lot of work and it is well thought out and he has some great insights that makes you pause and think. I just don’t think the data is good enough to build the fine detail relationships he has come up with. Just my opinion. But I respect other opinions as well because we all look at stuff through different lenses and filters.
Sorry for the ramble!
@golich428 I love rambles 🤩. You get a real sense of what people are thinking, and that can be beneficial to everyone. What I got out of Mr. B's book is that it gives you permission to enjoy (spend more money if you will) more during your early retirement by saying you can spend at a minimum 4.7% of your retirement account (given his constraints) not counting social security. This applies most to those whose retirement monies will be close to running out of money at end of life.
Up until recently I would not have called myself a Safety First guy. I have always been 80-90% US stocks (90% the past 15 years) which, for me, has been wildly successful (we both retired pre-60 years old). I started to change my outlook last October 2023 when US Treasury rates shot up and I started my Treasury bond ladder. Now with the changes caused by the present political stuff and likely resultant future changes (I fear they will be very bad), I am extending my Treasury ladder when the S&P 500 hits another record high, which is accruing frequently now. My bond ladder now extends from 2027 to 2035 and I will likely extend it out to 2040 this week. At that point we will both be getting social security. I think that counts as Safety First 🤗.
@golich428 @jkandell You are kind of losing me with the statistical discussion. I think you are making it more complicated then needed. You have all the data, historical stock and bond returns over the past 100 years, you have the inflation history, and the CAPE history. Given a starting condition similar to today (pick a year in the past that's similar to today) and see what happened over the next 30 years. That's it. Will it be a perfect fit, no. Markets are driven by human goofiness. But I think its good enough to make a retirement plan. Throw in a little safety factor and you are good to go.
Mr. B has a link to all his Tables and Figures in his book, many in color: https://www.bengenfs.com/charts-tables-for-you/
@pizzaman Bill’s process requires the retire to monitor the plan year by year with his curve of year by year withdrawal rates and make adjustments if it deviates too much - especially if it is caused by inflation.
Therefore he acknowledges that the starting SWR is not always reliable and it requires significant intervention in some cases. It is not a set it and forget it process like his earlier 4% rule was interpreted to be.
There is no way that two variables (inflation and CAPE10) explain how your retirement will unfold assuming all other variables are held constant. We have changed the way we measure inflation over time so it is not the same number. CAPE10 is somewhat arbitrary, it could be CAPE5 or CAPE15 at in given period that best fits the data. These are not static variables like in physics equations. 10lb mass is always 10lb mass in physics. He knows this because he is an MIT graduate.
I think he recognizes all this and has attempted to develop a process that takes these uncertainties into consideration. That is why some of his implementation examples are complicated - his words not mine. But, I think he has out run the data and folks might think it is again a rule rather than a process with much uncertainty that requires intervention at times.
However, you will probably be ok following his process - but do you have the will and mental capacity to know when those adjustments need to be made? In real time it is hard because you can’t see the next data point compared to looking back and saying “ I knew that”.
Setting a starting SWR based on worst case conditions is probably going to turn out ok but trying to set it based on current CAPE10 values and inflation ranges that historically have only happened a few times may not be very robust. When I replicated his study, I had plots of remaining portfolio values over the 30 year period that would test any retirees ability to “stick with the plan”.
In my opinion, it is unfortunate the 1990’s work he did was turned into a set it and forget it rule. I think part of his motivation is to use additional research to try and educate all of us that it is not that simple.
I have chosen a different path but I recognize other preferences exist. Some are total return and Bill’s work may appeal to them. Some are safety first and they probably cringe when they see his process. Others are probably some mix of both.
FYI - you may want to add a few books to your library: Burton Malkiel’s A random Walk Down Wallstreet and Wade Pfau’s Retirment Planning Guidebook just to name a few. Oh Charles Ellis Winning The Losers game is also good.
So I attempted to explain how I approach retirement. What is yours - anybody can chime in - this is just a friendly discussion on a topic I think everyone on this forum is somewhat passionate about - not just nuts and bolts about Pralana but about how we approach it when we turn from building a castle to dismantling it brick by brick.
t
