1Year Berkshire Price Prediction - Inflation +17% From 10/22

Stocks A to Z
62

@tedthedog , @johnIII -

I did regressed against ‘log(book)’. Here is my scientific reason: my data are still quite noisy / messy so I just clicked different regression options in excel until I found one that looked pretty.

I really, really, really like how @tedthedog laid out the analysis and data.

It’s odd how our P/BV data are different. I’m not sure what to make of it / fix it other than going thru all the quarterly reports and matching price data → that’s an immense amount of work and it seems just taking Morningstar P/BV data is a ‘good enough’ dataset to develop an understanding if Berkshire is a ‘good enough’ buy at the prevailing price.

As far as R - my experiences are in Minitab, PSPP, and SPSS. I’ve never used R; it looks like LinkedIn Learning has some decent ‘how to’ courses. Any other recommendations for learning R?

2 Likes
63

I definitely agree with this. Jim liked to endlessly tweak his models. I share some of that propensity. It’s kind of the 90% solution - You spend 10% of your time getting to the 90% solution and the remaining 90% of your time solving the last 10% of the problem. I think you have the 90% solution, regardless of the specific BV dataset you use. Likely the next obvious tweak would be to improve the BV estimates. As you say, that involves a lot of time. Jim offered a post on how to do that is this thread: OT bond funds - Mechanical Investing - Motley Fool Community. I think @DTBoojum also made BV estimates that he posted to these boards, and he’s still active. Perhaps he’d be willing to share? It would be interesting to see how the models might differ using different BV series. Probably not a lot, indicating the model has some robustness.

Anyway, I’m not throwing work on anyone’s plate :slight_smile: I really appreciate the work that has been shared here. I’ve been wanting to go through the same exercise but never got around to it.

6 Likes
64

I just posted a new topic about getting historical IV for Berkshire, and did so not having seen @cwags02 and @johnIII posts just about the same time that I posted.

In my post I asked if people have a time series for IV (not book).
What prompted my question is that my Morningstar data goes back to 1991.
I’ve done a somewhat similar analysis to @cwags02 with that longer history data, and things start to fall apart.

In tracking down ‘why??!’, one obvious glitch is the huge jump in book, and hence fall in P/Book that occurred in last quarter of 1998. So, what’s up wiith that?

From the 1998 annual report:
"
Our gain in net worth during 1998 was $25.9 billion, which increased the per-share book value of both our Class A and Class B stock by 48.3%. Over the last 34 years (that is, since present management took over) per-share book value has grown from $19 to $37,801, a rate of 24.7% compounded annually.

Normally, a gain of 48.3% would call for handsprings – but not this year. Remember Wagner, whose music has been described as better than it sounds? Well, Berkshire’s progress in 1998 – though more than satisfactory – was not as good as it looks. That’s because most of that 48.3% gain came from our issuing shares in acquisitions.

To explain: Our stock sells at a large premium over book value, which means that any issuing of shares we do – whether for cash or as consideration in a merger – instantly increases our per-share book-value figure, even though we’ve earned not a dime. What happens is that we get more per-share book value in such transactions than we give up. These transactions, however, do not deliver us any immediate gain in per-share intrinsic value, because in this respect what we give and what we get are roughly equal. And, as Charlie Munger, Berkshire’s Vice Chairman and my partner, and I can’t tell you too often (though you may feel that we try), it’s the per-share gain in intrinsic value that counts rather than the per-share gain in book value.
"
Ideally, we should be using IntrinsicValue not BookValue.
Or, perhaps someone with skills in fundametnal analysis has a suggestion about how to simply rejigger book?

Anyway, one lesson learned is that you should always the plot the time series of interest.
Look for any dislocations, then go back to the annual reports to see if they were related to corporate events.

5 Likes
65

R has, or at least had (maybe they’re better tutorials now), a hard learning curve. I manage to do what I need with it, but having worked before with true R experts, I know that I’m most definitely still an amateur. In fact, I should probably start over myself with a tutorial on the ‘tidyverse’ (an R thing) but I have invested the time yet.

I’m not worried about small discrepencies in reporting data values, or reporting trading day versus calendar day, in Morningstar data. But it’s good to be aware that such exist, i.e. that even Morningstar can’t be considered a total ‘gold standard’. In this case at least, the discrepencies are small.
I’ve also looked in detail at some CBOE stuff on options data. There were lots of discprencies, contradictions etc in that data, some of them significant, and this came right from the horse’s mouth i.e. CBOE.
Moral: treat all sources of financial data as suspect until proven otherwise.

Don’t know what to say about the ‘log’ thing.

4 Likes
66

The problem with that is that Intrinsic Value (IV) is subjective, so there’s not really a calculation for it. For instance, does WEB add to the IV? When he steps down, does that immediately change the IV? We can calculate BV, and then the hope is that IV tracks BV. E.g. IV = 1.5 BV, for instance. As you point out, BV can have discontinuous jumps, which messes with your model. I believe Jim dealt with some of those using one-time adjustments. For instance, I think that Jim viewed the tax breaks by the last administration (4 years ago?) as providing a one-time 4%(?) increase to BV. So all previous BVs could be increased by 4%(?) to smooth this out (I think that’s what he did).

Jim also believed that there was a change in the markets valuation of BRK following the credit crunch, maybe starting in 2012(?). So your model may be more robust if you stick to data subsequent to 2012(?).

Additional thoughts. Perhaps the best objective definition of IV is the current market price. Essentially everyone is placing their vote as to what the IV is and it is reflected in the price. If that doesn’t match your estimate of IV, then you of course can buy or sell, in effect placing your vote as to what IV should be.

4 Likes
67

This makes an immense amount of sense to me.

1 Like
68

@johnIII

Assuming I’m interpreting your comment correctly, I could not disagree more.

What happened to Mr. Market is irrational and market price has little to do with value?

Yes, price may squiggle around it, but typically with huge swings. The beauty of Jim’s analysis is that BRK price is so predictable based on IV (well, book as a proxy for IV, but as noted, book has some issues).

3 Likes
69

Well, notice I said “objective” definition. We can all agree on the current price, and it reflects everyone’s belief about what the IV is.

Like you I suspect, I believe BRK’s current price is lower than it’s intrinsic value. So I’m a buyer at these levels, casting my vote. But apparently everyone does not agree with us, as BRK is trading below out IV estimates. So, what would your definition of BRK’s IV be?

70

I agree it’s subjective, e.g. WEB said even he and Charlie could disagree on the number for IV (somewhat).
I’m not an expert on fundamental analysis and on intrinsic value, so I just reference below the Berkshire Owner’s manual for WEB’s discussion of IV (and to all the subsequent discussions).
https://www.berkshirehathaway.com/owners.html

3 Likes
71

I would suggest that the chart is actually two curves, one concave upward and one concave downward, meeting at the ends, like a slender football. The upper, concave downward curve corresponds to multi-year periods of increasing P/B, and the lower, concave upward curve corresponds to multi-year periods of decreasing P/B.

I once made similar plots for the S&P 500 index for forward 5-year and forward 10-year returns. I segregated the data into two sets, one for long periods (about 16 years) of rising P/E and one for long periods of declining P/E. These long periods of rising or falling P/E are obvious in Robert Shiller’s data, using either CAPE or trailing 12 month P/E. The two sets of data fit second order polynomials, one concave upward and one concave downward, meeting at the ends. Segregating the data into two sets, one for long periods of rising P/E and one for long periods of falling P/E, helped explain the wide scatter of forward returns near the center of the data set.

In the case of Berkshire Hathaway multi-year periods of rising or falling P/B are less obvious than the long periods of rising or falling P/E for the S&P 500, but to the degree that we can forecast whether P/B will rise or fall over the next 12 months, the better we can forecast the forward one-year return.

8 Likes