When the US Falls, So Does Korea — A Lead Correlation of 0.34, and Why It Can't Be Traded

This series has mostly confirmed that "a correlation looks like it exists but doesn't." This case is the opposite — the correlation clearly exists. And yet the conclusion is the same.

Method

• KOSPI daily returns vs S&P500 (SPY) and Nasdaq 100 (QQQ) daily returns, 2010-01 to 2026-08, 4,318 overlapping trading days
• Align the lag: the Korean market opens after the US market closes, so US prior day (t-1) → Korea same day (t) is the "lead." Matching on the same date (t, t) is in fact a misaligned pairing

Results

The lead correlation is twice the concurrent one. The regression is KOSPI ≈ 0.41 × S&P500 (prior day).

The larger the drop, the sharper the relationship becomes:

But This Is Not a Trading Signal

① It is already in the price. That the US fell -1% is known to everyone before the Korean market opens. The Korean market opens with a gap down by that much. The -0.81% in the table above mostly occurs at the open. Since you cannot have sold at the prior day's close, this correlation is not tradable information but a description of market structure.

② Explanatory power is 11%. R² 0.113 — 89% of the Korean market's movement is not explained by the US market.

A Real Case — September 2, 2026

Prior day US:  S&P500 -0.69%   Nasdaq 100 -1.27%
Same day KR:   KOSPI -3.89%    KOSDAQ -1.68%
Regression:    -0.28%

The decline the model explains is -0.28%; the actual was -3.89%. Most of that day's drop came not from the US market but from Korea-specific factors. The press cited Middle East risk, oil prices, interest rates and foreign investor selling as the reasons — those are after-the-fact explanations, not facts we tested. What we did confirm is one thing: that day's decline fell far outside the range explained by the US market correlation.

What This Result Says

The size of a correlation and its tradability are different questions. The signals rejected in this series (following investor flows, chasing spike stocks) failed because "the correlation is too small"; this relationship fails because "even with a large correlation, it is already reflected." To beat the market with public information, it has to be something others do not know, or something others cannot endure. The US market close is neither.

Data Sources

• KOSPI, SPY and QQQ daily closing prices. Long-term index trends can be checked on the market page