Look-Through Earnings Miscalculations

The challenge with look-through earnings over extended periods – and when there is trading – is
that the calculation becomes far more complex. If we are not careful it becomes easy to double-count.

Just as an example, when Berkshire sold PetroChina in 2007, it booked a capital gain – included in the reported EPS, of several $B. But did we already book some (unrealized) look-through while we held the position? Should we adjust?

Whether the holding is PetroChina, H&R Block, HCA or whatever – when we sell the position, we have an actual realization. In order not to double-count, we can take a couple of seat-of-the-pants approaches – but what we can’t do (without kidding ourselves) is add the look-through earnings to the EPS inclusive of capital gains.

One of the more interesting examples of this problem may be Gillette. When Berkshire shuffled Gillette off to P&G in 2005 in exchange for P&G shares, Berkshire booked a (non-cash) $5B gain that’s reflected in the earnings per share reported for 2005.

Meanwhile, in some ‘look-through plus reported earnings’ calculations we’re including both the look-through and that gain. Even if it were a cash gain, it would be double-counting. In this example, where the position is still effectively held (and look-through earnings still being 'booked' in look-through calculations - before and after the merge) the double-counting is more obvious.

Buffett hasn’t mentioned look-through in the annual reports in several years, as I'm aware, where he used to comment on them more regularly. Look-through works for understanding buy-and-hold-forever portfolios far better than it does for those with even some revolving positions. The calculation gets tricky, and maybe that's why we haven't see mention of it in a while.