Post New

I have always been fascinated by Mungofitch's posts where he will talk about BRK.B's current price in terms of Price/Value ratio, and report things like "at this P/V, the expected return in 1 year is 12% above inflation or -5% loss below inflation" or whatever.

I have also recently been fascinated by AI.

I decided to put these things together. I used chatGPT to help me write Python code that would help me analyze, and especially to visualize the historic performance of BRK.B stock st arting at various levels of Price/Value. The results are on blog here: https://kazart.blogspot.com/2024/09/predicting-ber...

A quick summary of what I did and what I found:

I used Book value per share as my Value metric.

I used Python to scrape 5 years of BRK.B price history from Yahoo financial website.

I created by hand using SEC filings a table of Book/Share values for BRK.B that goes back more than 5 years.

The data I got covers a range of P/B ratios from about 1.1 to over 1.7.

I show price and P/B vs time for 5 years. P/B over 5 years seems to have a trend upwards. I wonder what could be happening?

I found annual real stock returns from ~50% at P/B = 1.1 down to ~0% at P/B > 1.5

For stock hold periods less than a year, the results are a lot noisier, but there is still clear outperformance in stock return for P/B <= 1.3 even for 3 month holding periods.

There are no real large parts of parameter space where the real return on the stock is negative. This suggests it will may not be possible to consistently make money with Covered Call strategies.

Anyway, I think the plots I made from the data I processed are cool and evocative. If you are a visual learner and you are fascinated by the hint that the market may not be all that efficient, you will enjoy looking at these plots on the blog site https://kazart.blogspot.com/2024/09/predicting-ber...

R:)

5 years isn't long enough to derive any meaningful conclusions.

Here is a site that has P/B back to 2004. https://www.financecharts.com/stocks/BRK.B/value/p...
You said you derived it using SEC filings for more than 5 years, so why didn't you do more than 5?

The time between datapoints is not regular, which would make analyzing it a bit painful. Which is why I haven't done it yet.*

BTW, you don't really need Python to scrape price history from Yahoo financial. Cut/paste works just fine.


-------------------
[*] Duh, I just realized it would be easy. In Excel, just have another table of historical prices and do a table look up for each P/B datapoint. **smacks forehead**

Here is a site that has P/B back to 2004. https://www.financecharts.com/stocks/BRK.B/value/p...
You said you derived it using SEC filings for more than 5 years, so why didn't you do more than 5?

That's an odd little site. On the one hand, P/B back 20 years. On the other hand, it skips lots of quarters of book value reports.

On the third hand, they did a nice job of making it so you can't extract the data into a spreadsheet.

R:)

Here is book value data I have going back to the early 1990s. It's a merge of various datasets others have published over the years, so I can't guarantee its accuracy completely. But, I think it's fairly accurate, particularly for looking for general relationships. Use it if you find it helpful.

1991,"Q4",6440
1992,"Q1",6636
1992,"Q2",6990
1992,"Q3",7104
1992,"Q4",7743
1993,"Q1",7800
1993,"Q2",7887
1993,"Q3",8115
1993,"Q4",8853
1994,"Q1",8867
1994,"Q2",9158
1994,"Q3",9684
1994,"Q4",10081
1995,"Q1",11012
1995,"Q2",12313
1995,"Q3",13550
1995,"Q4",14019
1996,"Q1",14992
1996,"Q2",16528
1996,"Q3",17500
1996,"Q4",19011
1997,"Q1",19631
1997,"Q2",22732
1997,"Q3",22694
1997,"Q4",25488
1998,"Q1",28021
1998,"Q2",29743
1998,"Q3",24838
1998,"Q4",37780
1999,"Q1",38097
1999,"Q2",38189
1999,"Q3",36384
1999,"Q4",37987
2000,"Q1",37014
2000,"Q2",37760
2000,"Q3",38927
2000,"Q4",40442
2001,"Q1",38246
2001,"Q2",38454
2001,"Q3",37374
2001,"Q4",37920
2002,"Q1",39453
2002,"Q2",40659
2002,"Q3",40812
2002,"Q4",41727
2003,"Q1",42582
2003,"Q2",45986
2003,"Q3",46856
2003,"Q4",50498
2004,"Q1",51859
2004,"Q2",52304
2004,"Q3",52456
2004,"Q4",55824
2005,"Q1",56153
2005,"Q2",57205
2005,"Q3",58119
2005,"Q4",59377
2006,"Q1",61859
2006,"Q2",63300
2006,"Q3",66292
2006,"Q4",70281
2007,"Q1",71220
2007,"Q2",74499
2007,"Q3",77483
2007,"Q4",78009
2008,"Q1",77075
2008,"Q2",76176
2008,"Q3",77555
2008,"Q4",70330
2009,"Q1",66208
2009,"Q2",73812
2009,"Q3",81282
2009,"Q4",82052
2010,"Q1",89378
2010,"Q2",86650
2010,"Q3",90836
2010,"Q4",95433
2011,"Q1",97073
2011,"Q2",98709
2011,"Q3",96901
2011,"Q4",99837
2012,"Q1",106613
2012,"Q2",107356
2012,"Q3",111695
2012,"Q4",114196
2013,"Q1",120520
2013,"Q2",122873
2013,"Q3",126763
2013,"Q4",134986
2014,"Q1",138454
2014,"Q2",142483
2014,"Q3",144580
2014,"Q4",146226
2015,"Q1",146979
2015,"Q2",149743
2015,"Q3",151097
2015,"Q4",155495
2016,"Q1",157350
2016,"Q2",160012
2016,"Q3",163813
2016,"Q4",171565
2017,"Q1",178082
2017,"Q2",182843
2017,"Q3",187450
2017,"Q4",211718
2018,"Q1",211000
2018,"Q2",217654
2018,"Q3",228772
2018,"Q4",212500
2019,"Q1",224950
2019,"Q2",233997
2019,"Q3",243606
2019,"Q4",261417
2020,"Q1",229293
2020,"Q2",245765
2020,"Q3",264323
2020,"Q4",287031
2021,"Q1",293636
2021,"Q2",311276
2021,"Q3",316443
2021,"Q4",342622
2022,"Q1",345469
2022,"Q2",314090
2022,"Q3",310652
2022,"Q4",323637
2023,"Q1",345470
2023,"Q2",372996
2023,"Q3",363413
2023,"Q4",389367
2024,"Q1",397697
2024,"Q2",418800

Price to book has become increasingly less relevant due to steady buybacks. Buybacks occur at prices above book but below intrinsic value thus they increase the intrinsic value per share while simultaneously reducing book value per share. So the gap between intrinsic value and book value per share over time will become greater and greater. This is one of the main reasons Buffett had permanently abandoned growth in book value per share as a metric to measure progress of BRK.

I have P/B data for 61 arbitrary months from 10/2004 to 2/2023. 220 months.

The P/B ranges from 1.11 to 1.63.

I computed the gains for 6, 12, and 18 months for each of the 61 months, then looked at how the returns were distributed according to P/B.

P/B             6 mo.   12 mo.  18 mo.
Bucket: Count   Avg     Avg     Avg
1.10    3       9.80%   29.56%  44.54%
1.15    1      21.13%   31.87%  37.28%
1.20    6      11.75%   21.21%  33.34%
1.25    3       8.64%   10.51%  33.94%
1.30    7       7.96%   20.24%  21.21%
1.35   16       6.84%    7.65%  18.19%
1.40    7       6.93%    5.74%  11.86%
1.45    7       5.34%   16.04%  22.53%
1.50    6       3.24%    3.35%  -0.74%
1.55    5     -15.04%  -12.17% -11.43%

Read: There were 3 months where the P/B was between 1.10 and 1.15.

Yes, any conclusion is questionable since there were only 61 of the 220 months of data.

Yes, any conclusion is questionable since there were only 61 of the 220 months of data.

It's probably fair to say:

If valuation low forward return high.
If valuation high forward return low.

Right?

It's probably fair to say:

If valuation low forward return high.
If valuation high forward return low.
Right?

In short, yes.

I have data for all trading days since B shares were listed in 1996 (thank you old yahoo, curse you new yahoo). I also have book values taken from the quarterly reports for all quarters starting in 1996.

I choose to look at the P/B ratio going forward from December 2012 because it was drastically different after the BNSF acquisition and the 1.2X buyback threshold implemented around that time. You can pick dates within a couple of years of that and I don’t think you will get meaningfully different outcomes.

Take all dates from December 2012 forward and bucket them into P/B quartiles over this nearly 12 year period (<1.33x, <1.39x,< 1.44x, >1.44x). Now calculate the forward cagr from each date to today and place those returns in the P/B quartile buckets.

If you look at returns with a minimum 2yrs of history (stop at about October 2022 based on today’s date) you find median annual return rates by quartile of 16.9%, 14.7%, 13.8%, 12.9%. So yes, better forward returns if starting valuation is lower.

Incidentally I did a similar analysis with a stop date of November 2022 for the ending forward returns to avoid the current high valuations and found similar spreads on the forward returns of 12.4%, 11.1%, 10.6%, and 9.7%.

Jeff


As an aside:
We might think that the quartiles of P/B will change meaningfully over time like they did around 2011/12. I started tracking the forward quartiles from July 2018 when Warren removed the buyback threshold and up until about November 2022 they were significantly lower with median running about 1.33x. However, now that we have been experiencing higher than normal multiples for almost two years, the quartiles are nearly identical to the longer term ones. Maybe we will see significant change at some point but it doesn’t appear to be happening yet.

Price to book has become increasingly less relevant

Is there a published or relatively easily calculable alternative to book for estimating IV and/or its changes over time?

Thanks,
R:)

Is there a published or relatively easily calculable alternative to book for estimating IV and/or its changes over time?

Buffett did not give a formula for intrinsic value but he broadly outlined how to think about it. There are two main items that contribute to intrinsic value of BRK: (1) Net cash + investments per share, and (2) Normalized non-insurance operating earnings (excludes insurance u/w profit or loss, and investment income). This is commonly known as the "two column method". The first column is self explanatory and one can make an allowance for deferred taxes which may be at least partly payable at some point. The second column involves an estimate of business value of non-insurance subs run by Greg.

Post New

Berkshire Hathaway FAQ

---

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