The KOSPI crossed 7,000 this year. K-pop’s biggest agencies lost nearly half their value.
The good news everyone knows
In the last week of September, the KOSPI reclaimed 7,000, carried by chips and AI.
K-pop had a busy year too. BTS went back on stage. Big Bang and aespa returned with new music. For fans, it was about as good as it gets.
The stocks went the other way. As of September 1, HYBE (KRX: 352820), SM (KRX: 041510), JYP (KRX: 035900) and YG (KRX: 122870) were down 41% to 46% for the year. Over the same stretch, the KOSPI rose 62% (BigGo Finance).
Analysts have noticed. In April, one told The Korea Times that one-off, artist-driven events “seem to have limited impact on investor sentiment.” Nobody put a number on “limited.”
Comebacks, concerts, viral moments. Everyone calls them good news. Did any of it reach the share price?
We have the records to check.
A thermometer for attention
To see whether good news reached the stock, we first need to measure how big the news was.
Since January 2024, we have logged English Wikipedia pageviews every day for seven of K-pop’s biggest names: BTS, Blackpink, NewJeans, Le Sserafim, Stray Kids, aespa, and HYBE itself.
A pageview is one person looking a name up. It is not a fan count, and it is not album sales. It is a thermometer for public attention.
We call a day a spike when views reach at least twice the prior week’s average. We drop days when the usual level sits below 500 views; the noise is too high there.
The thermometer surprised us first. It was falling.
Average daily views across the seven names: 26,794 in 2024, 25,239 in 2025, and 23,850 in 2026 through September 22. That is 11% below 2024.
Attention may have moved to newer acts or new hits. So we do not say K-pop has cooled. We say the attention we measured fell.
Spikes still happened: 65 of them in 2.7 years. That raises the obvious question. What did the stocks do on those days?
65 spikes, graded
Test 1: When attention spikes, does the agency’s stock follow?
If attention lifts share prices, the artist’s agency should beat the market in the days after a spike.
The method is simple. We start from the close on the first trading day after the spike. Five trading days later, we check whether the agency beat the KOSPI. Measuring against the index keeps a rising market from passing as a fan effect.
The result: 29 out of 65. That is 45%. The median spike left the agency 0.5% behind the market.
Flip a coin 65 times and you get about the same. Against ordinary days, the result is indistinguishable from chance.
| After a spike | Beat the KOSPI | Median vs. KOSPI |
|---|---|---|
| 1 trading day | 48% | −0.3% |
| 3 trading days | 49% | −0.1% |
| 5 trading days | 45% | −0.5% |
| 10 trading days | 34% | −3.6% |
At no horizon did spikes beat the market more than half the time. The longer we looked, the worse it got.
Test 2: Is this just 2026?
This year, chips carried the market almost alone. Maybe K-pop was simply crowded out. So we checked week by week.
We asked whether weeks with rising attention were also better weeks for agency stocks. One number captures how closely two things move: 1 means lockstep, 0 means they ignore each other.
2024: 0.19. 2025: −0.16. 2026: 0.12. All three sit near zero.
Nothing broke this year. The two were never close.
Test 3: What about the biggest peaks?
Small spikes can be noise, so we isolated the ten largest. That is where the second surprise showed up.
Wall Street has an old line: buy the rumor, sell the news. Traders buy while good news is still expected. When it actually lands, they take profits. The day the news breaks often marks the top.
Our records follow that shape. Averaged across the ten biggest peaks, the agency’s stock beat the market by 2.7% in the ten trading days before the spike. In the ten days after, it trailed by 4.7%. Six of the ten followed this pattern.
One caveat. Six out of ten sits exactly on the threshold we set in advance. One different outcome would have flipped the verdict. We treat this as a clue, not a conclusion.
A side note
When artists at unlisted agencies spiked (Ateez, Ive, Rescene), the four listed agencies actually beat the market, 9 times out of 13.
Thirteen is too few to conclude anything. We log it and keep watching.
Across 2.7 years and 65 spikes, the agency beat the market 29 times. A coin flip does about as well.
The good news made headlines. The stock had moved first.
Conclusion
Across 2.7 years and 65 K-pop attention spikes, the artist’s agency beat the market five days later 29 times. That is indistinguishable from a coin flip.
This does not mean fan attention is meaningless. When the biggest wave of attention arrived, the expectations were already in the price.
So the thing to watch is not the charts. It is the days when earnings and tour revenue are reported, and whether attention arrives before the stock moves or after.
Limits
- We measured seven names on English Wikipedia only. Other languages, other names and new hits are not included.
- More than half of the 65 spikes came from artists at one company.
- Attention and share prices may have reacted to the same news at the same time. We do not claim attention moved the stock. We only report what the stock did after a spike.
- Chips drove the KOSPI this year, and every comparison against the index carries that effect.
What would prove us wrong
Starting from publication, we will track the next 10 spikes. If the median agency return five days later beats the KOSPI by more than 1%, and at least 7 of the 10 beat the market, we will withdraw our first conclusion. We will record the verdict here, with the date.
What we publish: aggregate figures, the spike definition, the decision rules, and the fingerprint of our calculation script (md5 d403dbeb…). What we keep private: the daily pageview series and per-company figures.
External figures: KOSPI 7,000 (Seoul Economic Daily, Sep 27, 2026); agency year-to-date returns (BigGo Finance, Sep 1, 2026); analyst quote (The Korea Times, Apr 28, 2026). Pageviews: Wikimedia public statistics.
Cited with attribution to Asia Pulse.
Since September 26, 2026, Asia Pulse has recorded a daily fingerprint of every paid output on X Layer: 0xb759Ea689FF49A6781232198fcAdD6fdfE0680cd