Hana Financial Group Inc. (KRX:086790) reported second-quarter net income of ₩1.21tn, down about 2% from the first quarter. On the surface, an unremarkable three months at a large Korean bank.
Underneath, two of the lines that matter most to a bank both moved hard and in the same direction. Credit loss provisions went from ₩230.1bn to ₩433.8bn, up 88.5%. Net interest income went from ₩2,505.3bn to ₩2,302.9bn, down 8.1%. Between them that's roughly ₩406bn of damage in a single quarter, on a business that earned ₩1.21tn.
Almost none of it reached the bottom line, and where the offset came from is the part worth understanding.
Set the ₩433.8bn against its own recent history. The eight quarters on file run ₩264.7bn, then unavailable, ₩297.2bn, ₩336.7bn, ₩289.2bn, ₩368.7bn, ₩230.1bn, ₩433.8bn. The first quarter of 2026 was the lowest reading in the series. The second was the highest, by a margin of ₩65bn over the next worst.
Going from the floor to the ceiling in one step is unusual enough to want an explanation, and the summary financials don't give one. Provisions can jump for several reasons that look identical from outside: a single large corporate exposure downgraded, a model recalibration that lifts expected loss across a portfolio, a management overlay taken deliberately in a good quarter, or genuine deterioration in retail or small business credit. The notes to the half-year report break the provision down by portfolio and stage, and anyone doing serious work on this needs to read them.
What the arithmetic does establish is scale. Provisions in the first half totalled ₩663.8bn against ₩633.9bn a year earlier, up only 4.7%. So on a six-month view nothing dramatic happened. The volatility is inside the half, not across it, which is more consistent with timing than with a credit cycle turning. That's the benign reading and it is probably the right one.
A cleaner way to see the pressure is to measure provisions against net interest income, which is the income they are set against.
Q2 2026: ₩433.8bn of ₩2,302.9bn, or 18.8%. Q1 2026: 9.2%. Q2 2025: 15.2%. FY2025 as a whole: 14.1%. FY2023, the year Korean banks absorbed real estate project finance losses and the group's provisions peaked at ₩1,550.7bn: 17.5%.
So on that measure the second quarter of 2026 was worse than the worst full year in the data. It is one quarter and the ratio is flattered on the wrong side by net interest income falling at the same time, which shrinks the denominator. Both effects push it up together, which is why the number looks so stark.
The interest income side deserves its own note. Interest income actually rose 4.1% sequentially, to ₩5,837.7bn. Interest expense rose 14.0%, to ₩3,534.8bn. That expense line had fallen for six consecutive quarters, from ₩3,840.1bn in Q3 2024 to ₩3,099.7bn in Q1 2026, as policy rates came down. It reversed sharply in the second quarter. Funding got more expensive faster than lending repriced, which is the classic squeeze, and it explains the entire decline in net interest income.
Now the arithmetic that most summaries skip.
Take the disclosed components of the second quarter: net interest income ₩2,302.9bn, plus net fee and commission income ₩893.5bn, less SG&A of ₩1,290.7bn, less provisions of ₩433.8bn. That totals ₩1,471.9bn. Reported operating income was ₩1,594.4bn. There is ₩122.5bn unaccounted for.
Run the same calculation on the first quarter: ₩2,505.3bn plus ₩771.2bn less ₩1,198.4bn less ₩230.1bn equals ₩1,848.1bn, against reported operating income of ₩1,653.6bn. That leaves negative ₩194.5bn.
So the residual line swung by roughly ₩317bn between the two quarters, from a substantial drag to a substantial contribution. That residual is other operating income: trading gains and losses, foreign exchange, securities valuation, dividends and the insurance result. The summary financials don't break it out.
For this bank in this quarter there's an obvious candidate. Hana runs the largest foreign exchange and trade finance franchise in Korea, a legacy of its origins as a currency-focused institution, and the won moved substantially over the second quarter. Two Samsung Electro-Mechanics contract filings, dated three weeks apart in the same period, converted dollars at ₩1,544.20 on June 29 and ₩1,475.60 on July 22, a 4.4% move. A bank with large FX positions makes or loses real money on moves like that.
That is inference rather than disclosure. What is not inference is that the second quarter's result was carried by a line item nobody follows, and that the same line was negative three months earlier.
The other offset is more durable. Net fee and commission income was ₩893.5bn in the second quarter, the highest in the data, up 15.9% from the first quarter and 29.9% from ₩687.7bn a year earlier.
That's not a one-off. Fee income has gone from ₩1,715.1bn in FY2022 to ₩2,719.0bn in FY2025, up 58.5% in three years, in a period when net interest income barely moved. A bank building a fee franchise while its spread business stagnates is diversifying in the direction it should.
Fee income is also less volatile than trading income and does not consume capital the way lending does, which matters given the group's stated commitment to hold its CET1 ratio above 13% while returning more than half of earnings.
The third-quarter provision, reported in late October. If it comes in near the ₩290bn to ₩370bn range that characterised most of 2025, the second quarter was a catch-up for an unusually light first quarter and the six-month total of ₩663.8bn was always the right number. If it lands near ₩430bn again, something in the loan book is deteriorating and the first quarter was the anomaly.
The second thing to watch is interest expense. One quarter of a 14% jump after six quarters of decline could be deposit competition, a shift in funding mix, or a rate move working through. Two quarters of it would mean the spread compression is structural, and a bank targeting 12% return on equity cannot get there while its core income line is shrinking.
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.