For fifteen straight years, the most common new Korean fund was a derivative. In 2024 that stopped.

Korea’s fund registry does one thing very well: it dates things. Every registered code carries the day the fund behind it was established. Nobody seems to read the file that way, so we did — 183,351 codes, sorted by the year they were set up and by what type they were.

Read like that, the shelf has changed hands three times.

Three eras on one shelf

The middle era is the striking one. For fifteen consecutive years — 2009 through 2023 — the single most common thing registered on the Korean fund shelf was a derivative fund, and in six of those years it was more than half of everything registered.

Then it stopped, in one year.

Derivative registrations fell 85 percent in a single year while the overall shelf shrank by 28 percent. That is not a slow retreat inside a shrinking industry; it is one category being switched off and another switched on in its place. Fund-of-funds codes rose from 604 to 893 to 1,281 across the same three years and are now nearly half of everything newly registered.

We can date the change precisely. We cannot say from this file why it happened — the registry records what was registered and when, not what anyone decided. Readers who follow Korean retail finance will know what else was happening in early 2024; this file neither confirms nor contradicts it, and we are not going to borrow the connection and print it as a finding.

The biggest cohort on the shelf is from 1998

The same read produces a second thing nobody counts. The largest single establishment year still sitting on the Korean register is not recent:

A 3.4-fold spike in one year is exactly the shape a data error makes, so we checked it before writing it down. If it were a placeholder date, the codes would pile onto one day. They do not: 1998 spreads across all twelve months, with the weight in October (3,304) and November (3,359). And 10,410 of the 13,925 are bond funds against three equity funds.

That is a real cohort, and its composition dates it — a bond-fund flood in the autumn after the 1997 crisis, still on the register twenty-eight years later.

It also corrects something we published ourselves. Our earlier read of this registry described new registrations as peaking around 2020 near 6,000 a year. Within the window that piece looked at, that was right. Across the whole file it is not: 1998 and 2007 are both more than twice 2020, and we should have said so. The decline that piece described is real and, if anything, understated — but the peak it named was the peak of its window, not of the data.

What a count of codes cannot tell you

Everything above counts codes, not funds and not money. One fund holds a separate code for each share class, so a fund sold in eight classes appears eight times, with eight establishment dates. We could strip class suffixes off the names to guess at funds, and we are not going to, because the guessing would be invisible in the result.

Older years are also counted only through their survivors. Codes are not retired cleanly, but they do disappear, so 1998’s 13,925 is what is left of 1998 — the year was larger than that. Every fall stated here is therefore a floor, not the whole drop.

And we cannot name asset managers from this file. The standard code comes in at least four families (KR5…, K55…, KRM…, K5V…), and the four digits that identify a manager in one family do not in another. We hold no table that turns those codes into names. That is a gap we can close by finding the mapping, not by reading company names out of fund titles — so for now it stays a gap.

The full year-by-year table, including every type in every year, is on the fund shelf page.