Post New

This is a subject that comes up from time to time.

Let's assume you own nothing but Berkshire shares and intend to keep it that way, and you intend to live forever. You want your income over time to be as high as possible consistent with a reasonable expectation that your after-inflation income will rise, not fall. How much do you sell for income? A particular version of the old "safe withdrawal rate" discussion.

I thought I'd share a simple plan that I concocted. It's based on the notion that you need reasonably steady real income for an indefinite length of time, while recognizing that market valuation multiples can be temporarily very high for some stretches, which messes up some systems.

Here's my simple suggestion:
(1) Calculate a smoothed book value for Berkshire. Once per year take the last four annual reports, calculate book per share for each year, and average the four numbers. No need to get fancier. No need for an inflation adjustment or looking up quarterly numbers. It doesn't matter if you accidentally skip doing the calculation one year, or do it every quarter, just do it "whenever you remember".

(2) Calculate how much of that smoothed book value you own. Your current number of shares times that smoothed book per share figure.

(3) Each quarter, calculate how many dollars worth of stock to liquidate: 2.5% of the number calculated above.

The interesting thing is that the income generated is pretty darned smooth, and worked well going into and coming out of the big stock growth bubble, price bubble, and valuation bust of 1998-2002.

If you had started this in March 1996 and kept it up to date:
Rolling one-year real cash raised would never have dropped more than -4% from its peak. Inflation adjusting the income, it would never have dropped more than -5.4%. Note that this covers the stretch of valuation levels going from very high to very low in the 1998-2000 stretch.

The portfolio's real value would have risen over time. Skipping the rapid growth of value in the late 1990s, the real value of the portfolio would have risen 1%/year in the stretch 2002 to date. Flattish in the earlier half, faster in the later half. In the same stretch, the rolling-year cash withdrawals would have increased inflation + 1.55%.

In any given year, the real withdrawals ranged from 2.6% of the real portfolio value at the start of the year, up to 8.3% of the starting value, but an average of about 5.7%.

However, this is not the same as saying a SWR of 5.7% is a good pick. The plan varies income: if the rate of growth of the firm slows a little, the income gently adjusts. If the growth rate really slows or starts falling you will start eating into the capital, but the effect will be very gradual and you should still be good for many decades.

It works particularly well if you occasionally say "I don't actually need that much money this quarter, I'll sell less than what it recommends".

Here's a table of the annual inflation-adjusted income generated. Scaled to 100k income in 2025.

1996  27999
1997  33808
1998  42712
1999  57144
2000  66127
2001  71739
2002  72988
2003  69121
2004  69311
2005  69507
2006  70650
2007  74533
2008  75760
2009  75117
2010  75112
2011  74257
2012  71855
2013  74497
2014  78025
2015  81914
2016  84732
2017  86309
2018  89635
2019  91021
2020  95509
2021  98408
2022  97203
2023  94935
2024  96099
2025 100000

Jim

Thanks for posting. That is very helpful.

Here's a table of the annual inflation-adjusted income generated. Scaled to 100k income in 2025.
...
2025 100000

PS, after the 2025-12-31 liquidation giving the fourth payment of the year and topping the real withdrawal up to $100k, you'd have 2.511 A shares left (more likely 37670 B shares), market value about $1.89 million after the sale. Just to give a sense of scale of the size of portfolio needed for the level of real annual income in that table.

Jim

Very interesting. Thank you for sharing!

"Calculate how much of that smoothed book value you own."

Thanks, Jim. Good plan. Base the withdrawal on the value of the shares you own rather than the price. Over time the price will vary around value, occasionally by a lot, as in the 1998-2000 stretch you cite, but the value will be more stable than the price.

Thanks, Jim. Good plan. Base the withdrawal on the value of the shares you own rather than the price. Over time the price will vary around value, occasionally by a lot, as in the 1998-2000 stretch you cite, but the value will be more stable than the price.

Yes, that's the idea.

There is a small hidden disadvantage, of course. (isn't there always?)

For any SWR type scheme at all, there is a trade-off between steadiness of income and getting a good average valuation level for what you're selling. Since you're selling a relatively stable dollar amount, you're selling somewhat fewer shares when prices are high and somewhat more shares when prices are low. This definitely isn't enough to make the plan not work, but the unavoidable trade-off is that selling more shares when they're expensive (and fewer when cheap) is very effective for that extra percentage point of long run return, but it makes the income extra variable. Sometimes a steady KISS is just fine.

Hence the comment that it works particularly well if you occasionally say "I don't need that much money this quarter" and sell fewer. And particularly particularly if that just happens to be a quarter that the share price is really low : )

Jim

This is trivially easy with Google Gemini as well.

I get the following

Berkshire Hathaway (BRK.B) Estimated Year-End Book Value Per Share

Fiscal Year Ending      Book Value Per Class B Share
December 31, 2025        ~$332.44
December 31, 2024        ~$301.04
December 31, 2023        ~$259.05
December 31, 2022        ~$215.40

Then I asked it to assume I had 10,000 shares and wanted to sell 2.5%

It came up with

Selling 2.5% of the $2,769,825 average book value of your position results in proceeds of $69,245.63.
To reach your target of $69,245.63 at the current market price of $489.25 (the closing price as of June 12, 2026), 
you would need to sell approximately 141.53 shares.
 

So this would generate around $70K x 4 = $280,000 in annual income off a portfolio value of $4,800,000 or roughly 5.8% - which is right at the average...


tecmo
...

“ you'd have 2.511 A shares left (more likely 37670 B shares), market value about $1.89 million after the sale. Just to give a sense of scale of the size of portfolio needed”

Darn, I’m a couple of tens of thousands of B shares short of the goal.

This is trivially easy with Google Gemini as well...

Indeed, very easy.
But it's wrong. Only a tiny amount in this case, 0.62%, but it's probably worth doing it yourself.

Gotta love LLMs.

Jim

“In any given year, the real withdrawals ranged from 2.6% of the real portfolio value at the start of the year, up to 8.3% of the starting value, but an average of about 5.7%.”

Very interesting data and new twist. We have mainly been living off Berkshire by occasionally trimming shares for 3 years or so, starting shortly after entering the draw down phase. I recall from your previous data that one can draw down *on average 6%/year in the past* and still maintain the principal. Fortunately, we live comfortably and continue modest withdrawal rate with a nice margin of safety. Even after a few years into retirement, I find it’s still challenging to “relax & spend more” even with growth in Berkshire principal and other assets.

Many Thanks, Jim, for sharing your thoughts, reasoning and data!

The LLM indicated an approximate value, and I thought that precision wasn't that important in the strategy?

tecmo
...

The LLM indicated an approximate value, and I thought that precision wasn't that important in the strategy?

In this case it was close, but only by coincidence. It could equally have been off by a mile, as long as the answer sounded statistically plausible. If it doesn't know what book value per share is/was or what that means, which of course it doesn't, it's not something you'd trust to do any research or calculation that matters.

For fun, I gave the response you received to a different LLM and asked if it was accurate. It responded that no, it was close, but not accurate, and then gave four different book values per B share, all wrong by a bit. (Perhaps by using average shares outstanding during the period rather than shares outstanding at the date of the statements? --admittedly that would be a fairly subtle error that even a somewhat attentive human analyst might make)

Jim

“ you'd have 2.511 A shares left (more likely 37670 B shares), market value about $1.89 million after the sale. Just to give a sense of scale of the size of portfolio needed”

Darn, I’m a couple of tens of thousands of B shares short of the goal.

All is not lost. 2.511 A shares would be about 3767 B shares. So put
those other tens of thousands of B shares in a sock drawer for the kids.

vez

Hi JimAI,

Is it easy to how this method would differ from selling 4.5% of shares each year in equal quarterly installments over the same time period?


thanks in advance

Is it easy to how this method would differ from selling 4.5% of shares each year in equal quarterly instalments over the same time period?

The main difference is that selling the same percentage of shares each time is WILDLY more variable income.
The "multiple of the smoothed book you own" method has a rolling-four-quarters drawdown of real income of -4% from peak.
For selling a constant percentage of your shares, it's -33% from peak. Though that is aggravated because of the late-1990s price spike, so everything looks lower compared to that for a while.

For a different metric: since 2002, what is the percentage change in rolling-year income compared to two years earlier?
For the "percentage of book owned" method the standard deviation of that number is 4.2%, for the constant percentage of shares owned it's 15.8%.

Jim

you'd have 2.511 A shares left (more likely 37670 B shares)...
...
All is not lost. 2.511 A shares would be about 3767 B shares. So put

Oh well. I'm a better investor than I am a typist. Maybe I'm trying to blend the two by adding a zero here and there...

Jim

This is intriguing. We're still a few years out from needing to live off of Berkshire. When we do, my original thought was to sell, say, 1.5% of shares per quarter, perhaps selling nothing if price is below a certain threshold. But this approach may be preferable.

This is intriguing. We're still a few years out from needing to live off of Berkshire. When we do, my original thought was to sell, say, 1.5% of shares per quarter, perhaps selling nothing if price is below a certain threshold. But this approach may be preferable.

It's a simple method, and I think it falls into the "good enough" category.

Though I think this is a better one
https://www.shrewdm.com/MB?pid=698430964

A shorter summary:
Estimate what your position is actually *worth*, not market value. For Berkshire this could be a fancy value metric, or as simple as "1.5 times book value per share".
Do this a few years of data, do an inflation adjustment on all the figures, and smooth them a bit so it doesn't squiggle up and down from year to year. I used a four year (actually 16 quarter) smoothing.
You end up with something like the pink line here https://www.stonewellfunds.com/SmoothedRealValuePe...
Then, each period, sell however many shares it takes to get the true value of your portfolio back to what it was the last time you did a withdrawal. That's it.

If Berkshire's trend rate of value growth slows, your withdrawals gradually decline, and vice versa.
But by construction, the true inflation-adjusted value of your portfolio never changes...you only ever spend the increase in value. Unlike most SWR schemes, you get to spend ALL of the value increase.

The very simplest approximation of that would be: Periodically calculate the four year average book per share, and sell the fraction of your shares that corresponds to the percentage rise in that number since last time. Using 16 quarterly figures and a quarterly withdrawal cycle is recommended but not necessary.

How does this compare to the system at the top of this thread? The year to year income is more variable...it will be a bit lower for a while after a bad bear market, for example. But this "constant portfolio value" system has no critical dependency on any particular long run average market valuation multiple, so you know precisely what the long run trend of your portfolio value will be: same as when you started.

Jim

There are many alternatives to living solely from Berkshire shares,
reducing dependence on a single stock. One very interesting and imaginative
plan was suggested for consideration by Jim some years ago. I cannot find the
original source, so what follows is my recollection of his idea.

Upon retirement place half of your retirement funds in Berkshire and half in a
Nasdaq 100 equal-weight ETF, such as QQQE. After calculating the amount to be
withdrawn each quarter (using your chosen SWR method), withdraw that amount
from whichever is larger, Berkshire or QQQE. The stronger carries the
load, while the laggard is given a rest and time to recover or catch up.

Looking at a chart comparing the two over time, there would be periods of
time when withdrawals were from Berkshire and periods when withdrawals were
from QQQE. Recently, of course, QQQE would be carrying the load.

My personal preference would be this plan. And all credit, of course, goes to Jim.

vez

Upon retirement place half of your retirement funds in Berkshire and half in a Nasdaq 100 equal-weight ETF, such as QQQE.

That is still too much concentrated. In tech.

Problem is, people want to get the returns of a concentrated portfolio without the associated volatility.

Can't happen.

That portion should be in a more broad index, like VTI or even VT. Or maybe SPY.

I view Berkshire more like a semi-actively managed index fund.

"Upon retirement place half of your retirement funds in Berkshire and half in a Nasdaq 100 equal-weight ETF, such as QQQE."


oh my, nuf said. On this board, remarkable.

Upon retirement place half of your retirement funds in Berkshire and half in a Nasdaq 100 equal-weight ETF, such as QQQE.
...
That is still too much concentrated. In tech.
Problem is, people want to get the returns of a concentrated portfolio without the associated volatility.

Well, yes and no. Something like QQQE is going to be price volatile, but volatility isn't risk, most especially in this context. The risk is very low, much lower than SPY or QQQ or most equity funds, since no position is over 1% of the assets. Company specific risk is extremely constrained. If the AI bubble were to burst badly and a half dozen firms went bankrupt, it would still only be 6% of the money. If you're in a cap weight fund, it's "which six?" becomes an existential question.

Admittedly there is the problem of "everything in a bubble" as in 1999. Or "everything went bankrupt", I suppose. But, little noticed at the time, real average earnings among the Nasdaq 100 set reverted [pretty much] right back to the prior trend after the 2000-2002 recession and bear market, aided a bit by periodic index reconstitution. On a larger time scale value just kept chugging along, with temporary dips in recessions that fully unwound after.

So, if your concerns are high returns and smooth returns and safe returns, once you realize that smooth isn't as important as you thought, you can still get high and safe.

Admittedly, a BIT of smoothness is good. I wouldn't want to have retired in 1999 owning nothing but a diversified basket Nasdaq 100 firms, even knowing that the value was still chugging along. Assuming one is over a certain age, arguably the best way to increase smoothness is just to put a certain portion of your money into an annuity. The older you get, the more sense this makes.

Jim

“ Here's my simple suggestion:
(1) Calculate a smoothed book value for Berkshire. Once per year take the last four annual reports, calculate book per share for each year, and average the four numbers. No need to get fancier. No need for an inflation adjustment or looking up quarterly numbers. It doesn't matter if you accidentally skip doing the calculation one year, or do it every quarter, just do it "whenever you remember".

(2) Calculate how much of that smoothed book value you own. Your current number of shares times that smoothed book per share figure.

(3) Each quarter, calculate how many dollars worth of stock to liquidate: 2.5% of the number calculated above.”

Thanks for developing and posting this, Jim. A simple, elegant system. I’d been messing around with the “withdraw a % of smoothed IV growth” idea. It works too, but is a lot more work. This is easy, I’m in.

Thanks for developing and posting this, Jim. A simple, elegant system.

Though the future is not the past, it's interesting to look at pushing Mr Buffett's system to its limit. For the last 20 years he has simply donated 5% of his current share count each July. (plus other smaller donations, but let's ignore those for the moment). As has been frequently reported, the market value of his shareholding has still risen despite the falling share count.

What's the limit?
If he had donated 7.5% of his shares each July instead of 5%, his portfolio would still be bigger in market value today, even after inflation. (only about 4% bigger, but still...)

Consider someone who used that 75% number but had sold those shares for income each July rather than giving them away. Their inflation-adjusted annual income would have been 6.0% - 9.2%* of the original real portfolio value, with the same inflation-adjusted portfolio value at the end. Simple average income 7.79%. That's in the rear view mirror; the future may be more modest.

Other than the trick of picking just the right percentage, you sure can't beat that scheme for simplicity.

Jim

* Note, I think the annual income could have been considerably more variable than that...it seems July has historically been a more predictable month that most in terms of valuation level, in fact the lowest variability of any month. The most variable were Feb,Mar,Apr, then May if you're wondering. I have no idea if that is entirely random, most likely so. But if I were to sell only once every year, I'd probably do it in July...if the pattern continues then it's a help, and if it's all random then it's as good as any!

For the last 20 years [Buffett] has simply donated 5% of his current share count each July.

Just to nitpick, he typically donates the shares on the last day of June and the foundations sell them early July.

Post New

Mechanical Investing FAQ

---

Contact Shrewd'm
Contact the developer of these message boards.