Clearly, there is a big disconnect between the metrics the firm is using to calculate book value and the above three metrics you mentioned. Are you missing anything that the firm considers relevant to calculate book value?

What multiple are you using on the look-through earnings? Assuming you use a single multiple for all their businesses (from railroad to energy to insurance to manufacturing and services, etc.), is that a valid approach?


I was trying to see the trend in ROE. The "ROE" I'm attempting above is not true return on equity. Underwriting earnings and cap gains are too lumpy. Smooth them out and they make little difference to the "ROE" trend, so I didn't include them. BV obviously includes underwriting earnings and cap gains, so don't compare my "ROE" numbers with BV growth.

Multiples of earnings are not relevant to ROE. I just want to sum the earnings. For the 5-grove IV estimate I do use a single multiple. Valid? Yes, for my purposes. I don't care what Berkshire invests in, I just want to target a certain minimum portfolio earnings yield.

My conclusion: "ROE" is trending down the last three years, hypothesis confirmed?

Yet,
Sell Apple at 35x = 2.85% earnings yield
Buy T-bills at 3.6% rate
ROE goes up? Hmmm.

A better conclusion: Berkshire is far too complicated for a simple electrical engineer and financial amateur.
Buy at <=1.3x book if I have funds. That's been satisfactory so far.
Sell when I need the money, hopefully when it's >1.5x book :-)