Earnings per share and dilution

Earnings per share is net income divided by the number of shares. A company earning $500m with 250 million shares reports EPS of $2.00.

Profit alone tells a shareholder little, because a company can double its earnings and leave its owners no better off if it doubled the share count to do so. EPS answers the question a shareholder is actually asking: how much of the profit belongs to the share held.

The share count used is a weighted average across the period rather than the count on the last day, so shares issued halfway through a year count for half of it. A company that issued heavily late in the year therefore shows a share count below what it will carry next year, and the dilution arrives in the following period.

Basic EPS divides by shares outstanding. Diluted EPS divides by that plus everything that could become a share: unexercised options, restricted stock, convertible bonds. Diluted is the lower and more conservative figure, and it is the one to use unless there is a reason not to.

The gap between them measures how much of the company existing holders stand to have transferred away. A gap of a few percent is ordinary. A gap in the teens, widening year on year, is a material cost that never appears as a cash outflow anywhere in the accounts.

A company buying back its own shares reduces the count, so EPS rises even with profit flat. The table shows three years of identical net income and a share count falling 4% a year.

EPS grows 8.5% across the period while the business earns exactly the same amount. That is not fraudulent and the remaining owners do hold a larger fraction of the company, but it is a different achievement from earning more, and describing it as earnings growth without qualification obscures which one happened.

The check takes seconds: look at net income and the share count separately over the same years. Profit flat with a falling count is arithmetic. Both moving in the right direction is not.

Companies frequently report an adjusted or non-GAAP earnings per share alongside the reported figure, removing items management considers unrepresentative: restructuring, acquisition costs, and very often stock-based compensation.

The adjustments are not audited and not standardised, so the same event can be excluded by one company and included by another. Excluding stock-based compensation is the most consequential, because it is a recurring cost of employing people that the adjusted figure treats as exceptional.

Where the two figures differ materially and persistently, the gap is the number to look at. A company whose adjusted EPS has exceeded its reported EPS every year for a decade is describing its ordinary costs as extraordinary.

At a company with multiple share classes, reported EPS may be expressed per share of a class other than the one quoted. Berkshire Hathaway reports on the Class A basis, and one Class A share converts to fifteen hundred Class B shares, so setting that EPS against a Class B price produces a P/E near zero.

Depositary shares carry the same mismatch for foreign listings, where per-share figures are per ordinary share and the price is per depositary share. Any per-share number is only meaningful beside a price on the same basis, and a ratio far outside the ordinary range is usually this rather than a valuation.

Reading EPS growth without checking the share count, which is the difference between a business earning more and a business divided among fewer holders.

Comparing an adjusted EPS from one company against a reported EPS from another.

Using basic EPS at a company that pays heavily in equity, where the diluted figure is the one that reflects what a share will actually own.

Comparing EPS between companies as though it were a valuation. A $10 EPS is not better than a $1 EPS; it reflects how the shares were split, and only the ratio to price means anything.

Treating a single quarter as a trend. Quarterly EPS carries seasonality, one-off items and the timing of buybacks all at once.