No. of Recommendations: 5
(1) The Book data is adjusted by CPI (that is, Book is divided by CPI) and then the WMA is calculated.
(2) The mean (price data) / mean( CPI adjusted WMA book data) should provide multiple to have two lines track
(3) For the variance, are you using close/WMA or close - WMA?
(4) You use the previous 20 years worth of daily data for variance?
(1) yup
(2) yup. This just shifts the WMA up or down so that it kind of aligns with the price
(3) realclose/realbookwma -1. (percentage error, in effect)
(4) yup.
Re #4, you could use any time frame. Berkshire's valuation had sort of a step change downwards around the credit crunch end 2007 right after the price spiked, which seems to have been pretty lasting so far. So using the history from the start of 2008 corresponds to the "modern era" of valuation. That would be a pretty reasonable choice. (almost the same, but my 20 years uses 2.5 years of the era when prices were higher).
As mentioned, the method I used was a bit more elaborate, but not meaningfully different. The same, but using two different value yardsticks. It goes like this:
(a) Book per share data is adjusted by CPI, then 16-quarter WMA is calculated to get BookWMA
(b) Scale the BookWMA to minimize error versus real price in the last 20 years to get ScaledBookWMA
(c) Value metric per share data is adjusted by CPI, then 16-quarter WMA is calculated to get ValueWMA
(d) Scale the ValueWMA to minimize error versus real price in the last 20 years to get ScaledValueWMA
(e) Take the simple average of ScaledBookWMA and ScaledValueWMA. That's what's on the graph.
So what is the value metric in (c)? The method I use is a bit complicated, but the difference versus using book value is (to date) so small that you needn't worry about it, you could just use book. The main difference is that is it a pinch more pessimistic lately because operating earnings have not been great lately.
In case anybody is interested, my valuation method goes like this:
* Calculate market value of all investments. Cash, equities, fixed income, equity method investments.
* Reduce the investments by 30% of insurance "float" (provisions for claims yet to be paid), to approximate the amount of investments which will forever earn nothing after tax and inflation -- insurers have to have a lot of ready cash on hand.
* Optional: Reduce the investments by the amount of any big stock positions trading over 21 times earnings (replace market value with 21 times earnings). Apple got pretty expensive for a while there, and it was a BIG position.
* Calculate the trailing-four-quarter net after-tax income from operating subsidiaries, being rails, insurance, and manufacturing/service/retail. These are stated at the start of the management discussion and analysis section of each quarterly and annual report.
* Adjust the operating earnings by replacing the actual trailing-four-quarter underwriting profit/loss with a cyclically adjusted figure. I use the average of (1.3% of float) and (2.7% of premiums earned).
* Multiply the operating earnings by a constant P/E to get some estimate of value of those businesses. I use a multiple of 15, but it isn't critical.
* Add the adjusted total investments from above to the proxy value of the operating earnings to get a total value for the firm.
* Divide by shares outstanding at end of reporting period to get value metric per share.
This result is not the true value of a share...what does that even mean?...but the idea is that it rises at the same rate as true (but unknowable) value per share. So, it can be used to look at the rate of increase of value, or to look at the historical relationship between price and value metric to get an idea of over/undervaluation relative to history.
This method has some quirks which are not obvious. For example, it ignores debt. I will continue to use the model only so long as I consider the debt level to be sustainable and in effect trivial. Another quirk is that it implicitly assumes that, if the cash pile is unusually large as it is right now, most of it will eventually get deployed as part of the normal average mix of portfolio allocations. i.e., it doesn't look at the current cash as the stuff earning nothing, it just assumes that an *average* amount of stuff will earn nothing over time.
Jim
