While that may be true for how you estimate intrinsic value, you also make adjustments to intrinsic value that should not be reflected in a marked to market view of share price. Apple holdings, for example, get discounted in your estimate, but often end up being (on a market price basis) more of a % of total BRK value than would make sense.

I agree that the method I use "should not be reflected in marked to market view of the share price". But I also assert pretty forcefully that it would be wilfully dumb to use a mark-to-market valuation for a position that large and volatile: the goal is to estimate value, not price.

For a comparatively minor position, it doesn't much matter what multiple the market is assigning: OXY is at a low multiple of earnings and Moody's is at a high one. There are lots of positions in the portfolio average, and no single one is really material enough to bother about doing a cyclical adjustment.

But both the market value and intrinsic value of the Apple position make up a lot of the value of a Berkshire share, around a fifth, so the accuracy of its valuation is material. I don't see anything magical about using the current market price. We know that can be a sensible number, or not at all sensible, depending on the weather. It's arguably the worst of the obvious ways to go about valuing the position.

Since it's too big to ignore, I just treat Apple as I would any operating subsidiary: a multiple of net earnings. I apply the occasional eyeball adjustment to the EPS if the current figure seems anomalous, but the progress is pretty steady. Consequently my value estimate goes up with the look-through earnings trajectory, rather than gyrating with the stock price. Since none of us thinks the position is about to be sold, that seems pretty reasonable. Who cares what the market value is?

I am not unwilling to stipulate that it's a particularly valuable franchise: I do use an earnings multiple that is 40% higher than the one I use for every [other] subsidiary, and 55% higher than Apple's own historical average P/E in the 10 years to 2019. That hardly seems to be an overly conservative valuation peg.

The earnings-based approach has quite a number of advantages, to my mind. Primarily that the metric is much steadier over time, which gives a more reliable overall estimate of the value of a Berkshire share over time. I'm much more interested in getting a steady metric than in getting the "right number", as my main uses for my valuation exercise are estimating the rate of change, and looking at the ratio of price-to-yardstick compared to the historical average of that ratio.

FWIW, I created a trend line of the log real value per share for each quarter since end 2002, and the current value is smack on trend (0.5% difference). That value-per-share trend line has risen inflation + 7.95%/year. But it's up only inflation + 6.1%/year in the last 4 years, having been a bit above trend for a while there.

Jim