Meritz Financial Group Inc. (KRX:138040) terminated a ₩700bn buyback trust on August 28 and signed a new ₩500bn one the same day. The pairing looks like continuity, and in one sense it is. But the reason the first contract ended matters: it had a full year to run, through March 25, 2027, and it stopped five months early because the money was gone. The stated purpose of termination in the filing is one line — acquisition through the trust contract is complete. Meritz spent ₩703.8bn, a shade more than the headline number, and closed the account.
That is the setup. The argument I want to make is narrower than "Meritz returns a lot of capital," which everyone already knows. It's this: the 50% promise is now being funded almost entirely with share repurchases rather than dividends, and the binding constraint on that program has quietly shifted from earnings to the holding company's own distributable-profit ceiling. The same filing that announced the new trust discloses exactly how much room is left. It's less than most people assume.
Meritz's mid-term policy, restated in the August 28 filing, is a total return ratio of 50% of consolidated net income, with the split between dividends and buybacks decided by comparing the yield on each. That last clause used to sound like boilerplate. It isn't. The company decided not to pay a year-end dividend on 2025 earnings and to route the entire return budget into buying and cancelling stock, on the reasoning that the repurchase yield exceeded the cost of the cash alternative.
You can see it in the cash flow statement rather than take it on faith. Meritz paid ₩241.1bn of dividends in FY2025 and ₩448.3bn in FY2024, both on a consolidated basis. In the first half of 2026 the line reads ₩1.3m — not billion, million, which is rounding noise. The treasury-acquisition limit table inside the August 28 trust filing tells the same story from the other side: dividends resolved at the ordinary general meeting since the prior fiscal year end, nil; interim and quarterly dividends resolved by the board, nil.
So the whole program is one instrument now. For a US reader used to financials that pay a steady quarterly dividend and buy back opportunistically on top, this is the inverse arrangement, and it is deliberate. The quarterly value-up filing on August 12 gives management's own capital-allocation test in a single number: an expected return on buying and cancelling stock of 16.4% against a required return of 10%. As long as that gap holds, cash goes to the repurchase line.
Add the two trusts together and the policy year running from the March 2026 shareholder meeting to the March 2027 one carries ₩1.2tn (~$860m at roughly ₩1,400 to the dollar, an approximate rate) of committed buying. Against a market capitalisation of ₩19.91tn at the August 27 close of ₩119,000, that is about 6% of the company.
Now hold it against earnings. Trailing-twelve-month net income through the second quarter is roughly ₩2.46tn — that's ₩668.4bn, ₩323.3bn, ₩680.2bn and ₩791.4bn for the four quarters from 3Q25 through 2Q26. Half of that is ₩1.23tn. The committed ₩1.2tn of buybacks is therefore not most of the 50% promise. It is essentially all of it, with the dividend contributing nothing.
The track record behind the promise is longer than one year. Between March 2023 and March 2026 Meritz repurchased and cancelled about 36.0m shares, roughly 17.7% of the shares outstanding, for around ₩3.9tn. Shares outstanding today are 167.3m. A company that has retired close to a fifth of itself in three years has earned the benefit of the doubt on execution, and the market has partly given it: the value-up filing reports cumulative total shareholder return of 156.0% since the policy was announced in 2023, an annualised 38.1% over three years.
Here is the part that doesn't get discussed. Korean law caps treasury purchases at distributable profit, and distributable profit is computed on the holding company's separate books, not the consolidated group. The August 28 filing spells the calculation out in millions of won as of the FY2025 close.
Net assets of ₩3,449.7bn, less paid-in capital of ₩113.1bn, less accumulated capital and earned reserves of ₩1,240.3bn, less unrealised gains of ₩26.2bn, gives a distributable-profit limit of ₩2,070.2bn. Against that, ₩1,091.0bn of treasury stock has already been acquired since the end of FY2025. Remaining capacity: ₩979.2bn.
The new ₩500bn trust consumes roughly half of what's left. That still leaves headroom, and the number resets when FY2026 earnings are booked into the separate accounts. But it reframes the question. Consolidated equity is ₩12.79tn as of the second quarter, and it would be easy to look at that figure and assume the buyback runway is effectively unlimited. It isn't. The runway is a ₩2.07tn separate-books number that gets refilled once a year by dividends the subsidiaries push up to the holding company, and a bit under half of this year's refill is already spent.
The stock trades at about 8.1x trailing earnings and 1.56x book, against a 52-week range of ₩100,600 to ₩146,200. It sits well above the Korean bank holding companies on price-to-book — KB, Shinhan, Hana and Woori are the natural comparison set on a screen, and Meritz's return on equity of roughly 21% on FY2025 net income of ₩2,350.1bn against average equity of about ₩11.1tn is the reason for the premium.
But 8x earnings is still 8x. Management's 16.4% expected return on repurchase is close to the arithmetic inverse of a low-teens multiple adjusted for the growth they assume, and it's the honest reason the dividend went to zero. Buying stock at 8x earns more than handing the same cash to shareholders who are taxed on it. If the multiple re-rated toward the high teens, that calculus would flip, and by the company's own stated framework the mix should shift back toward cash. Nobody should assume the all-buyback structure is permanent. It's conditional, and management has said so.
The obvious one is that the earnings funding all this are more cyclical than a bank's. Meritz Securities carried real estate finance exposure of about ₩9.6tn at the end of 2025, roughly 127% of its own equity, against a large-brokerage average nearer 59%. Assets classified as precautionary-or-below at the securities arm rose from ₩882.7bn at the end of 2023 to ₩1,969.6bn at the end of 2025. A credit cycle that forces provisions would cut net income, and 50% of a smaller number is a smaller buyback. The policy has no floor in won.
Second, the distributable-profit ceiling can tighten rather than loosen. It depends on subsidiary dividends reaching the holding company, and insurance subsidiaries operate under capital rules that can restrict upstreaming when solvency ratios move.
Third, and this cuts against the bears as much as the bulls: buying back stock at 1.56x book reduces book value per share even as it lifts earnings per share. That's a defensible trade at a 21% ROE. It stops being defensible if returns normalise toward the mid-teens and the multiple doesn't fall with them.
Two things settle this. The first is the board resolution due around March 2027, which will cancel the shares bought under both trusts and reveal the actual share-count reduction rather than the announced won amounts. The second is the size of the trust signed after the March 2027 shareholder meeting. If FY2026 net income lands near an annualised ₩2.9tn and the next policy year's commitment comes in materially below ₩1.45tn, the constraint isn't willingness. It's the ceiling in that filing table.
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.