The Faucet and the Drain: What a Buyback Headline Doesn't Tell You
Every quarter, dozens of companies announce that they "returned capital to shareholders" through stock buybacks. The number is always big, it is always in the first paragraph of the press release, and it is usually said in the same breath as the dividend, as if the two were the same thing.
They aren't the same thing, and this piece is about the difference. We aren't against buybacks. By the end you will see a company that does them about as well as they can be done. But a buyback headline leaves out two things a dividend cannot leave out, and once you see them you will read those press releases differently.
We pulled the numbers for fourteen of the fifteen stocks we own, straight from each company's annual report to the Securities and Exchange Commission (the filing is called a 10-K), and we added one outside company as a yardstick. The fifteenth, Bank OZK, reports to bank regulators instead, and we say so below. No estimates and no analyst opinions, with one approximation we flag where it appears.
The faucet and the drain
Think of a company's share count as the water level in a bathtub. Lower water is what you want: fewer shares, so each one you hold is a bigger slice of the business.
A buyback is the drain. The company spends cash to buy its own shares and retire them, the share count falls, and every remaining share owns a little more. That is the part the press release talks about.
What the press release doesn't mention is the faucet. Nearly every large company pays part of its executives' and employees' compensation in stock: new shares, created out of nothing, handed out every year. The filings call it stock-based compensation. We will call it stock comp. That faucet runs the whole time the drain is open, and some of the buyback, sometimes most of it, does nothing except mop up the shares the company just gave away.
Here is what that looked like at Nike over its last five fiscal years, 2022 through 2026.
Nike spent $16.9 billion buying back its own stock over those five years. In the same years it recorded $3.6 billion of stock comp expense, the accounting value of the stock and options it granted to executives and employees. The share count fell from 1,611 million to 1,481 million. That is an 8% drop.
Stock comp expense equaled roughly 21 cents of every buyback dollar. That is a rough measure, not an exact one (the expense is the value of grants when they are made, not the market value of shares on the day they land), but it is the closest thing every company reports the same way, and it says how much of the buyback offset new grants rather than shrinking the company. The headline never mentions it.
One thing to be fair about: stock pay is still pay. Had Nike paid those people in cash, the money would have left the company anyway. The point is not that the grants are improper. The point is that the headline counts the buyback that offsets them as a return to you.
The second thing the headline leaves out: timing
Stock comp is the faucet. The other problem is when companies usually choose to open the drain.
| Nike fiscal year (ends May) | Buybacks | Avg. share price |
|---|---|---|
| 2022* | $4.0 billion | $145 |
| 2023 | $5.5 billion | $112 |
| 2024 | $4.3 billion | $102 |
| 2025 | $3.0 billion | $75 |
| 2026 | $0.15 billion | $63 |
*Partial year: price data begins October 2021.
Nike did the bulk of its buying at its highest prices, and almost none once the shares got cheap. In fiscal 2023 it spent $5.5 billion at an average price of $112. In fiscal 2026, with the average price down to $63 and a low near $42, it spent $146 million, about two weeks' worth of the old pace. As this is written, on October 13, 2026, the stock is about $35 and the buyback has all but stopped. Put another way, the $16.9 billion Nike spent is roughly a third of what the whole company is worth at that price.
That is not a Nike quirk. It is how buybacks often go, because a buyback is discretionary. It is usually among the first things cut when business turns down, which is exactly when the shares are cheapest. A dividend works the other way. Cutting a dividend is public and embarrassing, so companies defend it. Nike raised its per-share dividend in every one of those five years, including the bad ones. That lopsidedness is the whole reason a long dividend streak means something and a buyback announcement means much less.
How it is supposed to work: Berkshire Hathaway
We added one company we don't own, as a yardstick, because it does the thing the press releases describe.
Berkshire Hathaway spent $47 billion on buybacks from 2021 through 2024. Its stock comp expense over the same four years was zero. Not small. Zero. Its filings report no stock-based compensation expense and state that Berkshire has no equity compensation plans; according to its proxy statements, its managers are paid in cash tied to the results of the businesses they run. Every dollar of that $47 billion retired a share, and the share count fell by roughly 7% (our one approximation, from Berkshire's reported Class A equivalent shares).
And look at the timing. Berkshire bought $27 billion of stock in 2021, when, by its own stated policy, it judged the shares cheap. It bought $3 billion in 2024 as the price rose. In 2025, with the stock near record highs for much of the year, it bought none. The rule, as described in its 10-K since 2018, is that Berkshire buys back only when the price is below what the company conservatively thinks the business is worth. The drain opens when the water is cheap and closes when it isn't. That is a buyback run for the owners.
Berkshire also pays no dividend, and that makes our point from the other side. Its buybacks are as good as buybacks get, and the only way a shareholder can know that is to read the filings. A dividend needs no filing. It shows up in your account.
Every stock we own, from the filings
Five fiscal years for each company, except Berkshire (four, through 2024) and Bank OZK, which files its reports with bank regulators rather than the SEC, so its figures are not in the database we used. Dollars in billions. The table is sorted by the fourth column; percentages use unrounded figures.
| Company | Years | Buybacks | Comp ÷ buybacks | Comp ÷ profit | Shares |
|---|---|---|---|---|---|
| Berkshire (yardstick) | 2021 to 2024 | $47.0 | 0% | 0.0% | about −7% |
| Lowe's | 2022 to 2026 | $37.5 | 3% | 3.1% | −20% |
| Zoetis | 2021 to 2025 | $8.5 | 4% | 2.9% | −7% |
| Tractor Supply | 2021 to 2025 | $3.0 | 9% | 4.9% | −8% |
| PPG | 2021 to 2025 | $2.0 | 12% | 3.7% | −5% |
| EOG Resources | 2021 to 2025 | $7.0 | 13% | 2.8% | −7% |
| Comcast | 2021 to 2025 | $45.5 | 14% | 9.1% | −20% |
| ADP | 2022 to 2026 | $7.7 | 15% | 6.3% | −4% |
| Medtronic | 2022 to 2026 | $9.6 | 21% | 9.1% | −5% |
| Nike | 2022 to 2026 | $16.9 | 21% | 15.6% | −8% |
| PepsiCo | 2021 to 2025 | $4.6 | 36% | 3.9% | −1% |
| T. Rowe Price | 2021 to 2025 | $3.2 | 40% | 12.1% | −4% |
| McCormick | 2021 to 2025 | $0.2 | 166% | 7.7% | 0% |
| NextEra Energy | 2021 to 2025 | none | n/a | 2.5% | +5% |
| Realty Income | 2021 to 2025 | none | n/a | 4.4% | +119% |
Three groups fall out of that table.
Companies whose buybacks genuinely shrink the share count. Lowe's spent $37.5 billion against a faucet of only $1.1 billion, and its share count fell 20%. Zoetis, Tractor Supply, PPG, EOG, Comcast, and ADP are in the same family. Lowe's is the closest thing we own to the Berkshire model on the faucet. On timing it is not: its heaviest buying came in the years ending January 2022 and 2023, at average prices of about $238 and $201, and it nearly stopped in the year ending January 2026 at an average of about $240, before the stock fell to the $180s this autumn. Its slowdown was not about price.
Companies whose buybacks mostly tread water. At Nike, Medtronic, PepsiCo, and T. Rowe Price, stock comp equaled between a fifth and two-fifths of the buyback dollars. PepsiCo's number looks worse than it is, because PepsiCo barely buys back stock at all; it sends seven times as much out as dividends. T. Rowe Price is the one that gives us pause, with the second-highest stock comp as a share of profit in the portfolio.
Companies that barely buy back stock, or issue instead. McCormick barely buys back stock at all. NextEra Energy and Realty Income go the other way on purpose: they issue new shares to fund power plants and properties, their share counts go up, and their dividends go up anyway. Realty Income's share count more than doubled in five years through stock-funded acquisitions and share issuance, and it has raised its dividend for more than 30 consecutive years, by the company's count. The share count was never the point. The dividend is.
The number we think matters most
The ratio of stock comp to buybacks is the one that makes the headline, but it isn't the fairest number. It punishes companies that barely buy back stock and flatters companies that buy back a lot. The fairer question is the one in the fifth column: of every dollar the company earns, how much is paid out as new stock to its executives and employees?
For Berkshire it is zero. For NextEra, EOG, Zoetis, Lowe's, PPG, and PepsiCo it is under 4 cents. For Nike it is almost 16 cents, and for T. Rowe Price 12. Nike's figure is pushed up by two years of collapsed profits as well as by the grants themselves, which is worth knowing, but those two are the outliers in our portfolio, and if we were scoring this today they would be the two we would look at hardest.
We are considering adding this figure to every stock write-up on this site, and possibly to the scoring itself. We haven't decided, and when we do we will publish the rule before we apply it, the same way we published the five boxes. For now, it is here so you can see it.
What to do with this
You don't need to read 10-K filings to protect yourself. Two habits are enough.
First, when a company announces a buyback, look for the share count. It is at the bottom of the income statement in every quarterly report, in the line that says "weighted average diluted shares" (the average number of shares out there during the period, counting the ones promised to employees). If the buybacks are real, that number goes down, year after year. If it doesn't go down, new share issuance offset the buyback. One wrinkle: that diluted count also shrinks a little on its own when a stock falls, because fewer employee options are worth exercising, so a falling stock flatters the number slightly.
Second, prefer the return you can see. A dividend lands in your account on a date the company announced in advance, in an amount you can check to the penny. We track every one of ours on the portfolio page, down to the 43 cents Realty Income sent us in July. No buyback has ever shown up there, and no buyback ever will.
Data: annual 10-K filings for each company, fiscal years 2021 to 2025 (2022 to 2026 for companies whose fiscal year does not end in December; Lowe's years are labeled by their ending date in January, not by Lowe's own fiscal-year naming), retrieved from the SEC's EDGAR XBRL database. "Stock comp" is the stock-based compensation expense each company reports. "Buybacks" is cash paid to repurchase common stock. "Profit" is reported net income. Share count change uses weighted average diluted shares, adjusted for Tractor Supply's 2024 split. Berkshire's share count change is approximate, from its Class A equivalent share count. Nike and Lowe's average share prices are the mean of daily closes in each fiscal year, from Nasdaq's published price history; the earliest year for each is partial. This is not financial advice. Nothing here is a recommendation to buy or sell any security. We own every stock in the table except Berkshire Hathaway, and we may buy or sell any of these holdings at any time.