Two filings landed within hours of each other on August 28. Meritz Financial Group Inc. (KRX:138040) terminated the ₩700bn buyback trust it signed in March, then signed a new ₩500bn one running to March 31, 2027. Read the headlines and nothing happened: a company that buys its own stock is going to keep buying its own stock.
Read the footnotes and something did happen. Buried in the termination filing, under the section for other matters relevant to investment decisions, is a sentence that changes how and when Meritz's repurchases turn into a smaller company. Until now, the company cancelled treasury shares immediately when a trust contract closed. From the shares acquired under this program onward, it will cancel once a year, by board resolution around March. The shares bought in the trust that just ended are scheduled for cancellation in March 2027.
That's the story. The buying continues at roughly the same pace. The share count does not.
The numbers in the termination filing are specific. Before termination, Meritz held 6,273,560 common shares acquired within the distributable-profit range — 3.75% of the company. Under the old practice, those shares would have been retired on or about August 28 and shares outstanding would have dropped from 167.3m to roughly 161.1m. Instead they sit in treasury, and the outstanding count stays where it is until the board acts next March.
The 2026 movement schedule shows how quickly this machine normally ran. The trust started the year holding 4,366,000 shares, acquired 9,782,060 more, and cancelled 7,874,500 — all inside eight months. Cancellation used to follow acquisition by weeks. Now it will follow by up to a year.
For anyone computing per-share figures, that's a real distortion with a known expiry date. Earnings per share, book value per share, and any screen that divides a group number by shares outstanding will use a denominator through the winter that is 3.75% larger than the economics justify. Treasury shares carry no dividend and no vote, so the economic ownership has already transferred to remaining holders. The reported arithmetic just hasn't caught up.
Meritz attributes the change to a Commercial Act amendment dated March 6, 2026, which the filing describes as making cancellation of acquired treasury stock mandatory within one year. The company is, in other words, moving to an annual batch process because the law now sets an annual outer limit.
There's an irony here worth stating plainly. Korean treasury stock reform exists because companies historically parked repurchased shares indefinitely and later used them for cross-shareholdings, defensive alliances and management entrenchment rather than retiring them. Mandating cancellation within a year is a floor designed to stop that. Meritz was already well above the floor — it cancelled on the spot. Applying the new rule as the operating standard means moving from immediate to annual. Strictly on promptness, that's a step backward, prompted by a rule intended to be a step forward.
I don't think it signals bad faith, and it would be silly to argue otherwise given the track record: about 36.0m shares, roughly 17.7% of the company, bought and cancelled between March 2023 and March 2026. Batching is administratively simpler, aligns the cancellation with the March shareholder meeting and the annual policy cycle, and the one-year statutory deadline makes quietly holding the shares essentially impossible. But it does mean the visible evidence of the program — the share count going down — now arrives once a year instead of continuously, and a program whose credibility rests on visibility has given up some of it for convenience.
There's a second-order question the filing doesn't answer. The new trust runs to March 31, 2027, which is around or just after the expected cancellation resolution. Shares bought in the final weeks of that contract may well miss the March 2027 batch and wait for March 2028. The filing doesn't say. It's a detail, but for a company this precise about its return program, it's an odd gap.
The other thing the new contract makes explicit is that Meritz commits money, not ownership. The filing states the ₩500bn trust is expected to buy 4,201,681 common shares, and then immediately notes the calculation uses the ₩119,000 close of August 27, the day before the board met, and that the actual count will vary with the price.
Work through what that means. At ₩119,000, ₩500bn retires about 2.51% of the company. At the 52-week high of ₩146,200 it retires roughly 3.4m shares, about 2.0%. At the 52-week low of ₩100,600 it retires nearly 5.0m, close to 3.0%. The spread between those outcomes is more than a full percentage point of the company, decided entirely by where the stock trades over the next seven months.
This is the standard mechanic of a fixed-currency buyback and it cuts against the shareholder when the thesis works. A rising price means the same won buys fewer shares, so the accretion per share falls exactly as sentiment improves. Meritz's own capital allocation framework — an expected repurchase return of 16.4% against a 10% required return, per the August 12 value-up filing — is a function of the price paid. That framework implies the program should scale down if the multiple expands. Sizing the trust in won rather than shares is at least consistent with that logic, even if it's the less flattering outcome for anyone who wants a guaranteed share-count reduction.
The obvious objection is that I'm making too much of a scheduling change. Cash spent is cash spent, the shares are contractually destined for cancellation, and a seven-month delay in the bookkeeping changes no cash flow. That's fair, and if the March 2027 resolution lands on schedule and cancels the full 6,273,560 shares plus whatever the new trust buys, this will read as a non-event in hindsight.
The less comfortable objection runs the other way. A company holding 8,694,662 treasury shares in total — the 6,273,560 plus 2,421,102 of fractional shares left over from the past split and the 2023 comprehensive share swap — has an asset it did not have under the old regime. Nothing in the filing suggests any intention other than cancellation. But the whole point of prompt cancellation as a governance signal is that it removes the option rather than asking investors to trust that it won't be used.
Third, the termination itself was triggered by exhaustion, not by a change of heart. The old trust spent ₩703.8bn against a ₩700bn mandate and closed five months early because it was full. Nothing about the August 28 pair suggests the program is slowing.
One date and one number. The date is the board meeting expected around March 2027, and what to check is whether the cancellation resolution covers the full 6,273,560 shares from the closed trust plus everything the new trust accumulates, or only the first block. The number is shares outstanding in the annual report filed next spring. If it comes in near 161m, the mechanism worked as described. If the count is still close to 167m, the change was more than administrative.
kstock reads DART every morning and writes up what moved — the contract, the buyback, the number that does not add up. The daily post and a Saturday roundup, by email.