A Korean ETF above its stated value falls back tomorrow — unless it holds Asian stocks, in which case it rises

There is a piece of advice that circulates whenever retail investors discover exchange-traded funds: check the premium. If the fund trades above its net asset value, you are paying more than the holdings are worth, and you should wait.

Across every Korean ETF that traded between January 2024 and August 2026, that advice is correct — and then, for one group of funds, it inverts completely.

What a premium usually predicts

Take every fund-day where an ETF closed in the top 5% of its own historical premium range, and measure what that fund did the following day against its own average.

The first two rows are what the textbook describes. A fund that closed expensive relative to its holdings gives back about half a point the next day; a fund that closed cheap gains about the same. The correction is symmetric, it is large relative to a daily move, and it rests on 24,300 observations.

The last row runs the other way. An Asia-tracking synthetic fund that closed at a premium went on to outperform by a third of a point. Buying it at what looked like the worst moment was, on average, better than buying it on a quiet day.

We should be plain about the sample: 632 fund-days is a twentieth of the domestic sample, and the Asian physical row sits between the two extremes rather than alongside the synthetic one. This is a strong pattern in a small group, not a law.

The obvious explanation is wrong

The intuitive reading is that Chinese markets are shut while Korea trades, so the published asset value is stale and the ETF price is simply ahead of it.

The US row rules that out. American markets are not merely shut during Korean hours — they are shut entirely, with no overlap at all. If a closed underlying market broke the relationship, US-tracking funds would be the most broken of the lot. Instead they revert at −0.480 and +0.474, indistinguishable from domestic funds.

Whatever separates the Asian funds, it is not the simple fact of a closed exchange. We do not know what it is. The feed publishes each fund’s net asset value but not when that value was struck or against which session, and we are not going to invent a mechanism we cannot see in the data.

How large the deviations get

The prediction result above is about direction. Size is a separate question, and it splits along two axes that turn out to be roughly additive.

Being synthetic roughly doubles the typical gap. Holding foreign assets roughly doubles it again. Leverage, which is the feature that attracts the most warnings, adds the least: leveraged domestic funds average 0.390% against 0.289% for unleveraged ones, a difference of a tenth of a point.

At the extremes the range is wide enough to matter to anyone holding these funds:

The gap between the top and bottom of that list is a factor of roughly 300. A Korean money-market ETF tracks its stated value to a hundredth of a percent. A leveraged synthetic China fund has closed as much as 31.9% below its stated value and 18.2% above it — a span of fifty points on the same instrument.

Across all 605,072 fund-days, the average absolute gap is 0.414%, and 9.79% of fund-days exceed a full percentage point.

What follows from this

Not a trading rule. We measured closing prices, and the differences here are smaller than the spreads and costs on exactly the funds where they are largest.

What does follow is narrower and more durable: the premium figure is not one measurement. On the 1,068 domestic and US funds in this data it behaves the way the textbooks say and reverts within a day. On 22 Asia-tracking synthetic funds it does not revert at all, and reading it the same way would have pointed the wrong direction for two and a half years.

Anyone quoting a single premium number for “Korean ETFs” is averaging together two populations that do not behave alike, and the smaller one is the one where the number is largest and most likely to be quoted.