Korea lists bonds on three boards. Everyone quotes one of them

There are three places a bond can be listed in Seoul, and they do not behave alike.

The one everyone quotes is the government board — the inter-dealer market for treasury paper. On 9 September it held ten listings. Across 22 sessions it never held more than twelve or fewer than eight. That handful of listings is what “the Korean bond yield” means in almost every sentence written about it.

Beside it sits the general board: 271 listings on the same day, ranging from 270 to 346 over the window. Bank paper, capital-company paper, utility paper, corporate paper.

And beside that sits a third board almost nobody outside Korea has counted: the small-lot board, 40 listings, and it is 40 nearly every day. Twenty-one of our 22 sessions show 40; one shows 39. It is a small, near-fixed universe of housing-bond series, and it is the board a Korean household actually buys from.

The three boards, 9 September

The general board sits 0.767 percentage points above government paper. The small-lot board sits 0.599 above it. The three boards are kept side by side, updated each trading day, on our bond boards page.

The gap is not a constant

Over 22 sessions the general-board premium over government paper ran from 0.412 to 0.950 percentage points. The small-lot premium ran from 0.426 to 0.774.

A single number quoted as “the spread” is therefore a number picked from a range twice its own width. The government board moved too — its median ran 3.630 to 4.011 percent inside the same window — so the premium widens and narrows for reasons on both sides of it.

Sorted by who issued the paper

The board a bond trades on is not the same question as who owes the money. Sorting the same 9 September session by issuer type, using the listing names, gives a second ordering.

Financial-sector paper pays 0.663 percentage points over treasury paper. Housing bonds pay 0.343 below it, which is the one line in the table that deserves a second look rather than a conclusion — these are short, amortising series and the yield convention on them is not the treasury convention.

The 96 unclassified listings are unclassified because our sorting reads issuer type out of the listing name, and 96 names do not announce it. They are not folded into “corporate” to make the table tidier. A sorting rule that guesses would move a quarter of the board on our say-so.

What is missing from this, and why

Three columns arrive in this feed that are not used above.

The opening, high and low yield fields are unusable. Across 3,547 rows we found 22 with a negative high yield, and one listing whose high and low imply a same-day range of 23,602 percentage points. A ranking built on intraday range would be a ranking of feed errors, so those columns are left alone.

The residual-maturity and classification columns arrive filled on 2.8 percent of rows. No maturity bucket appears in this piece for that reason. Maturity is recoverable from listing names — a name carries its redemption month — and that is done separately. Where the trading actually happens is a third question again: on this board of a few hundred listings, the ten most-traded take 99 percent of the money.

And there was a hole. This series stopped after 24 August and stayed stopped for 18 days, because the dataset sat outside the check that tells us when a daily file is missing. The collector was never broken; nobody ran it. The 12 absent sessions were refetched from source on 11 September, which is why this piece has 22 sessions rather than ten, and the check now covers this dataset.