Since May 2026, every KOSPI company with assets of KRW 2 trillion or more has had to file English versions of its major disclosures, and the largest companies must do so on the same day as the Korean filing. For many overseas investors, the English filing is the only version they read. So we asked a narrow question: do the figures in the English filings match the Korean originals?
What we compared
- 235 English disclosures published on the Korea Exchange's English KIND between January and October 2026, each paired with its Korean original on DART.
- 6,870 Korean figures (amounts, share counts, ratios and dates) were looked up in the matching English filing by our own comparison program.
- The program flagged 7 spots where a figure or its unit did not match. On 3 November 2026 a person checked each one against the original filings on DART and English KIND.
What we found
Five of the seven flagged spots were real differences, and checking them by hand turned up one more that the program had missed. All six sat in two of the 235 disclosures, both treasury-share filings with long tables.
| Type | What the English filing showed | How it shows |
|---|---|---|
| Missing digit | A 13-digit won amount lost one digit and read as about one-ninth of the Korean figure | The English table's total no longer follows from its own lines |
| Swapped digits | Two adjacent digits in a share count were swapped | The row no longer matches the subtotal right below it |
| Changed digit | One digit changed in a total share count | The total row does not add up |
| Changed digit | The last digit of a computed acquisition limit was off by one | The limit does not follow from the lines above it |
| Unit label | A table headed "Amount (KRW)" whose values were in KRW million | Net assets read a million times smaller than they are |
| Column placement | Rows of trust-held shares placed under closing balance instead of share retirement | Same numbers, different meaning (found by hand) |
The other two flagged spots were not differences: the English filing had converted amounts from KRW million to full won, which is correct.
One pattern worth knowing
In every case, the Korean table added up and the English one did not. Re-adding the English tables would have caught five of the six differences without reading a word of Korean. Only the unit label needs a side-by-side look at the original.
For an analyst working from the English filing, that suggests a cheap habit: before using a share count or an amount from an English table, check that the table's own totals hold. We turned this and six other checks into a checklist for investors.
Update: a wider check of 1,153 disclosures
On 4 November 2026 we extended the same check to 28 large KOSPI companies: 1,153 English disclosures published on English KIND from January to October 2026, each paired with its Korean original on DART. A person checked every spot the program flagged against the original filings.
- 10 confirmed differences in 6 disclosures, about 0.5% of the disclosures in this sample. They include the six described above.
- Types: one changed digit (5), a missing or added digit that makes an amount read about one-ninth or ten times its size (2), swapped digits (1), a unit label (1) and a column placement (1).
- Where: 4 of the 6 were treasury-share filings (acquisition, disposal, trust contracts); the others were a share-exchange decision and a corrected dividend decision.
- Not differences: 4 flagged spots came from corrected filings that print the before and after values together. Compare corrected filings value by value after the correction.
What this does not show
- It is not a market-wide rate. Both samples (235 and 1,153 disclosures) come from large KOSPI companies over ten months. We do not draw a percentage for the market from them.
- Our program can miss things. The column-placement case was found by a person, not by the program. We have since added a check for it.
- We compare figures only. We do not judge the quality of the English sentences. One label in the same filing described a share price as an "estimated acquisition amount"; that is outside this count.
- Filings can be corrected. We checked the versions available on 3 November 2026.
- The program also lists lower-priority items, such as figures that appear on one side only, often because the English leaves out a note. We did not review those here.
Why we do not name companies
The point is the pattern, not the companies. We leave out company names and exact figures so the differences cannot be traced to a filing from this note.
Sources
- Korean filings: Financial Supervisory Service DART
- English filings: Korea Exchange English KIND
This note describes differences flagged by a comparison program and checked by a person against public filings. It is not a judgment on any company's disclosure practice.