I Thought My KOSPI 200 ETF Was Diversified. Over Half Was Samsung and SK hynix

Why owning several ETFs may not mean you are diversified

From October 12 to 16, Korea’s stock market was badly shaken.

The KOSPI fell 14.5% in a single week, and circuit breakers tripped on the 15th. Samsung Electronics and SK hynix, which had led the market’s rise, dropped 17.7% and 24.3%, more than the index itself.

The crash raises one important question for anyone invested in Korean stocks.

Were we really as diversified as we thought?

Buying a KOSPI 200 ETF means spreading your money across 200 of Korea’s leading companies. It is easy to assume the risk is spread more evenly than with individual stocks.

Yet the holdings of KOSPI 200 ETFs as of October 16, 2026 tell a different story. Just two stocks, Samsung Electronics and SK hynix, make up about 59% of the money invested.

You own 200 companies, but more than half of your money sits in two of them. And this crash showed what that concentration can lead to.

Which ETFs were retail investors holding? In the first half of the year, the top retail net purchases included SK hynix leverage, AI chip and KOSPI 200 ETFs. In September, interest shifted to US index and dividend or covered-call ETFs.

We looked at the holdings of 13 of these ETFs and how they fared in the crash.

The KOSPI 200, dividend and covered-call ETFs stand out. Their names and goals differ, yet all had close to 60% of real exposure to Samsung Electronics and SK hynix, and their crash-week returns were alike too.

Covered-call ETFs do use an options strategy and pay distributions differently. But a different strategy did not necessarily mean a different risk.

DIVA Quantizer’s correlation analysis points the same way. The correlation of monthly returns between KODEX 200 and TIGER Dividend Covered Call Active was 0.99: their prices have moved almost in lockstep.

Buy different ETFs, and the risk you actually carry may still be much the same.

The problem becomes clearer when you look at a whole portfolio rather than single ETFs. Suppose that at the end of September you held the following.

· KODEX 200: 40%
· TIGER Semiconductor TOP10: 20%
· TIGER Dividend Covered Call Active: 20%
· TIGER US S&P500: 20%

Korea’s benchmark index, chips, dividends and US stocks: the money is split across four ETFs. On the surface, it looks like reasonable diversification.

But running DIVA’s Markowitz model on data through September gave a surprising result: 96.4% of the portfolio’s total risk came from the three Korean ETFs.

The chip ETF, with a 20% weight, accounted for 30.1% of the total risk. The US S&P500 ETF, with the same weight, accounted for just 3.6%.

The reason was high correlation among the ETFs. The chip ETF and the dividend covered-call ETF had correlations of 0.92 and 0.98 with KODEX 200. Different products, but largely moving in the same direction.

This reveals an important distinction: the share of your money and the share of your risk are not the same.

Putting 20% of your money into a fund does not mean taking on only 20% of the risk. An ETF that is volatile and moves like your other holdings can take up far more risk than its weight.

So what matters in judging diversification is not the number of ETFs, but how independently they actually move.

Can the overlapping risk be reduced? This time we kept KODEX 200 at 40% and replaced the chip ETF and the dividend covered-call ETF with US Nasdaq 100 and physical gold ETFs.

Both had relatively low correlations with the Korean ETFs. Re-analyzing the new portfolio, annualized volatility fell from 37.7% to 22.6%.

Going a step further, we applied DIVA’s HRP model, which sets weights based on how alike the assets are and how risky each is. KODEX 200’s weight fell to 7.4%, and the portfolio’s estimated volatility dropped to 14.1%.

Did it make a difference in the actual crash? We assumed each portfolio was bought at the September 30 close and never rebalanced, then compared how they did in the October crash.

Despite being split across four ETFs, the original portfolio lost about as much as KODEX 200 alone.

The portfolio with lower-correlation assets cut the loss roughly in half, and the one rebuilt with HRP to spread risk further fell comparatively little.

Cutting risk had a price. Expected annual returns estimated from pre-crash data were 47.0% for the original portfolio, 35.8% for the swapped one and 25.9% for the HRP portfolio.

The portfolio that promised the highest returns in the rally took the biggest loss in the crash.

The point is not which portfolio was right. It is whether you knew how much risk you were taking for higher expected returns, and where that risk came from.

This comparison is an after-the-fact test of hypothetical portfolios computed with data through September. The assets were chosen for illustration; it does not mean the model predicted the crash.

Investing in the KOSPI 200 does not mean you hold the Korean market’s risk in even slices.

Adding a chip ETF can enlarge the risk you already hold, and a “dividend” or “covered call” label does not make a fund move differently from your stocks.

That does not make concentrating on a stock or sector wrong. You may concentrate risk on purpose in pursuit of higher returns.

The problem is believing you are diversified while actually holding the same risk over and over.

That gap is what DIVA Quantizer uncovered here. Correlation found the overlapping assets, risk contribution showed where the portfolio’s risk was concentrated, and the model calculated how the risk structure changed when assets were swapped or reweighted.

Next time you consider adding an ETF, rather than asking only whether it is something new, ask this:

“Does this ETF add a new opportunity to my portfolio, or am I buying a risk I already own one more time?”

Diversification starts not with collecting ETFs with different names, but with designing different risks.

Methodology and caveats

Sources: yfinance, Naver Finance ETF data (WiseReport), Koscom CHECK Expert+ (Money Today), ETF Check (Business Watch). Analysis script: backend/scripts/etf_samsung_hynix_analysis.py.

This article was written with the help of AI.