Korean investors are walking away from stocks, and the bitcoin “kimchi premium” is back in the headlines. The obvious story is that one is feeding the other. Two months of half-hourly records say it is not.
A bad week in Seoul
On September 14 the KOSPI fell 3.26% to 6,684. Foreign investors sold for a fourth straight session — 3.9 trillion won that day alone. SK hynix dropped 6.35%, Samsung Electronics 4.05%. Banks were among the few things that rose. Two days later the US Federal Reserve raised rates by a quarter point, its first increase since July 2023, with oil above $100 a barrel.
People are also leaving. Average daily trading on the KOSPI fell to 21.4 trillion won in September, less than half of June’s 50.3 trillion. The cash Koreans keep in brokerage accounts to buy stocks shrank by more than 40 trillion won in two months. On investor forums, one line keeps coming back: not trading stocks is better for your mental health.
In the same month, a different headline appeared. From September 1, bitcoin on Upbit, Korea’s largest exchange, was quoted about 1% above the price abroad — the “kimchi premium,” back after a quiet summer.
So the story writes itself: the money that left stocks went into crypto.
It is a reasonable guess. The press put the two events side by side; it did not check whether one moved when the other did. That takes a record kept since before either happened. We have one.
What the premium measures
The kimchi premium is the extra Koreans pay for the same coin. Most Koreans can only buy crypto with won, on Korean exchanges, and cannot easily move money to exchanges abroad. When many people in Korea want to buy at once, the local price runs ahead of the world price.
Think of a limited-edition sneaker that sells above retail only in Seoul’s shops. The size of the markup tells you how long the queue is. In 2022 the bitcoin markup reached 21.5%.
One limit up front. The premium is a thermometer for how urgently people in Korea want to buy. It is not a ledger of whose money moved where, and this article can only go as far as the thermometer goes.
What we recorded
Since July 9 we have logged the premium every 30 minutes for eight coins: bitcoin, ether and a group of meme coins such as Dogecoin and Pepe. Nothing was filled in after the fact. Over the same period we recorded Korean stock and sector closes every day.
One detail matters for everything below. We price the premium against what a dollar-pegged coin (USDT) costs inside Korea, because that is the route money actually takes to leave the country. Many headline figures use the official exchange rate instead. We come back to that difference at the end.
We ran three checks, and wrote them down before computing anything: is the premium really higher than usual; does it rise after the stock market crashes; and does the urgency spread to meme coins, where small investors crowd first.
Check 1. Is the premium really back?
A headline quotes one day. One day of 1% can be chance. So we first set the usual range from July 15 to August 31, then asked where September sits.
September’s median bitcoin premium was 0.00% — the exact middle of the summer. About one reading in four sat above the summer’s upper quartile, which is what you expect by chance. Across the whole record, nine readings out of ten sat within a tenth of a percent of zero.
The chart says one thing: nothing changes on September 1. Measured our way, the premium never left. If the money that left stocks had rushed into bitcoin, this is where it would show first. It does not.
Check 2. When stocks crashed, did bitcoin get more expensive in Seoul?
If money moves from stocks to crypto, it should move in order. A crash day comes first; some of the people who sold head to a crypto exchange; the premium rises in the two days after. More often than on an ordinary day.
The KOSPI fell 2% or more on twelve days in our record. Two of them fell on gaps in our log, which leaves ten. For each, we took the premium at the 3:30 p.m. close and again 48 hours later.
| Crash day | KOSPI | Bitcoin premium, next 48h | Meme coins |
|---|---|---|---|
| Jul 13 | −8.95% | down | down |
| Jul 20 | −4.46% | down | up |
| Jul 28 | −10.84% | down | up |
| Jul 29 | −5.98% | up | up |
| Aug 3 | −5.12% | down | down |
| Aug 6 | −4.58% | up | down |
| Aug 19 | −5.80% | down | down |
| Aug 24 | −3.12% | down | down |
| Sep 2 | −3.99% | up | down |
| Sep 14 | −3.26% | up | down |
| Crash days | 10 | 4 of 10 up | 3 of 10 up |
| Ordinary days | 31 | 48% up | 48% up |
After a crash, the bitcoin premium rose four times out of ten. On an ordinary day it rose 48% of the time. The typical move after a crash was slightly down, and only two of the ten beat the top quarter of ordinary days. A coin flip would have done about as well as “crash, then crypto.”
The biggest crashes are the clearest. On July 28 the KOSPI lost 10.84%; the premium edged down. On July 13, −8.95%; down again. If panic sellers were queueing at crypto exchanges, those were the days to see it.
Check 3. Did the urgency reach meme coins?
When small investors crowd in, cheap and jumpy coins heat up before bitcoin does — Dogecoin, Shiba Inu, Pepe. So if this were retail money on the move, the markup on meme coins should have grown more than bitcoin’s. If only bitcoin moved, we would suspect a few large buyers instead.
Neither happened. After crash days, the meme-coin premium rose only three times in ten, against 48% on ordinary days. It moved less than on a normal day, not more. On September 2 and September 14 — the two crashes the press wrote about — meme-coin premiums fell.
Conclusion
Seoul’s stock trading has nearly halved since early July, but the bitcoin premium in our record did not move with it: September sat in the middle of the summer’s range, and after ten crash days it rose less often than on an ordinary day, in bitcoin and in meme coins alike. The simplest reading is that Korea’s small investors did not take their money to crypto; they took it off the table. If you want to know when they come back, the number to watch is not the kimchi premium but brokerage cash balances starting to grow again.
That leaves the headline 1%. Priced against the dollar-pegged coin inside Korea, bitcoin carried almost no premium. So whatever gap the headlines saw sat mostly in what Koreans pay for dollars, not in what they pay for bitcoin. A premium on the dollar says something about the won. It says little about a rush into crypto.
Two things happening in the same month does not make one the cause of the other. Telling them apart takes a record started before either began, with stocks and coins in the same ledger. That is why we keep both in one.
The money left the stock market. It did not show up in bitcoin’s price in Seoul.
What would prove us wrong
One number.
- The next time the KOSPI falls 2% or more, if the bitcoin premium rises more than 0.09 percentage points within 48 hours — the top quarter of ordinary days — and meme-coin premiums rise with it, “the money did not go to crypto” was wrong for that day. Three such days in a row and we withdraw the conclusion.
Whichever happens goes into the log that day. Not rewritten later.
What this article cannot say
The premium measures urgency, not flows. We cannot see who sold stocks or who bought coins. Our record starts on July 9, so it holds ten usable crash days; that is enough to rule out a strong link, not a weak one. Two crash windows fell on gaps in our log and were excluded. One meme coin had missing Korean quotes and was left out of the meme group. We did not archive the won price of the dollar-pegged coin itself, so the split of the headline 1% is an inference from our method, not a measurement.
Press figures: Seoul Economic Daily (Sep 14), Bloomingbit (Sep 16), US Federal Reserve (Sep 16), Bitcoin Magazine and Bloomberg (Sep 1). Early-July and September trading values: Korea public data portal. Everything else: Asia Pulse, logged every 30 minutes; the three checks and the crash-day rule were written down before the numbers were computed and were not changed to fit them. Half-hourly series and per-coin readings are not published here. Crypto prices: data provided by CoinGecko. Cite with attribution to Asia Pulse.